1 00:00:01,760 --> 00:00:03,679 Join us as we gather around the hedge, 2 00:00:03,679 --> 00:00:05,139 where we dig into technology, 3 00:00:05,440 --> 00:00:08,160 business, and culture with the finest minds in 4 00:00:08,160 --> 00:00:09,139 computer networking. 5 00:00:20,894 --> 00:00:22,195 Well, hello, Tom. 6 00:00:22,574 --> 00:00:24,175 See, I said Yvonne I I started with 7 00:00:24,175 --> 00:00:25,774 Yvonne last time, so I'll start with you 8 00:00:25,774 --> 00:00:27,535 this time. Yeah. I switched it up. Hello, 9 00:00:27,535 --> 00:00:28,035 Russ. 10 00:00:28,750 --> 00:00:31,010 And now we cannot say happy New Year. 11 00:00:32,109 --> 00:00:34,929 Why is that? Because we talked earlier. 12 00:00:35,469 --> 00:00:37,149 Oh, is it still is it not still 13 00:00:37,149 --> 00:00:38,609 happy? I'm still happy. 14 00:00:40,350 --> 00:00:42,929 It's really not Happy how about happy Friday? 15 00:00:43,454 --> 00:00:46,674 Okay. Happy Friday. For a long weekend. Yeah. 16 00:00:46,734 --> 00:00:48,994 We we already do that. That'll work. 17 00:00:49,534 --> 00:00:51,314 And hello to Tom's plant, 18 00:00:52,255 --> 00:00:54,655 who is I I it's still back there 19 00:00:54,655 --> 00:00:55,155 typing. 20 00:00:55,695 --> 00:00:56,195 And 21 00:00:57,679 --> 00:00:58,899 Little Shop of Horrors. 22 00:00:59,759 --> 00:01:01,539 It's not the Little Shop of Horrors. 23 00:01:02,159 --> 00:01:03,379 It's a nice plant. 24 00:01:04,079 --> 00:01:06,000 What's the name of the plant in that 25 00:01:06,000 --> 00:01:06,500 movie? 26 00:01:06,879 --> 00:01:08,560 Oh, I can't remember. We're gonna have to 27 00:01:08,560 --> 00:01:09,780 go find that now. 28 00:01:10,125 --> 00:01:12,444 So hello, Yvonne. I hear you back there 29 00:01:12,444 --> 00:01:13,665 in the frog lamp. 30 00:01:14,125 --> 00:01:15,025 I am here. 31 00:01:15,405 --> 00:01:17,325 Happy Friday. And then we were joined by 32 00:01:17,325 --> 00:01:18,305 Mike Bushong, 33 00:01:19,805 --> 00:01:20,944 who I mean, Thanksgiving. 34 00:01:21,325 --> 00:01:23,105 I thought if we're just allowed to say 35 00:01:24,204 --> 00:01:25,185 picking a random 36 00:01:25,680 --> 00:01:26,739 a random, 37 00:01:27,519 --> 00:01:29,280 day or holiday that's, like, way in the 38 00:01:29,280 --> 00:01:30,959 past. If we're gonna do happy new year 39 00:01:30,959 --> 00:01:33,620 this late new year, happy thanksgiving to you. 40 00:01:36,560 --> 00:01:39,120 See, on our last recording, Yvonne dared to 41 00:01:39,120 --> 00:01:41,140 challenge Russ on this very point. 42 00:01:43,125 --> 00:01:44,344 Oh my goodness. 43 00:01:46,005 --> 00:01:48,105 So let's talk about constrained 44 00:01:48,484 --> 00:01:50,265 supply or constrained 45 00:01:50,724 --> 00:01:51,865 source environments 46 00:01:52,564 --> 00:01:54,745 because that is where we live today. 47 00:01:55,069 --> 00:01:57,649 The price of memory is going up. Crucial, 48 00:01:57,709 --> 00:01:59,170 I think it was, said 49 00:01:59,549 --> 00:02:02,769 no more PC market for us. No more 50 00:02:02,989 --> 00:02:03,810 no more, 51 00:02:05,069 --> 00:02:06,450 oh, there it is. Anyway, 52 00:02:06,750 --> 00:02:09,229 so, Audrey, that's now the name of your 53 00:02:09,229 --> 00:02:09,729 plant. 54 00:02:10,375 --> 00:02:10,875 Audrey. 55 00:02:11,814 --> 00:02:13,655 Okay. Good to know. So Perfect. 56 00:02:14,055 --> 00:02:15,675 Anyway It's settled. It's settled. 57 00:02:18,295 --> 00:02:20,455 So crucial, I think it is said that 58 00:02:20,455 --> 00:02:22,855 they're no longer gonna do PCs. They're not 59 00:02:22,855 --> 00:02:25,435 doing consumer market any longer. 60 00:02:25,879 --> 00:02:28,520 It's all gonna be server side AI side 61 00:02:28,520 --> 00:02:30,060 because the demand is too high, 62 00:02:30,520 --> 00:02:32,460 which is driving up the price of phones. 63 00:02:32,840 --> 00:02:34,860 It's driving up the price of PCs. 64 00:02:35,800 --> 00:02:38,040 It's actually getting hard to find things. In 65 00:02:38,040 --> 00:02:40,294 fact, there was a really kind of 66 00:02:40,594 --> 00:02:42,594 a funny thing I saw out on x 67 00:02:42,594 --> 00:02:45,155 where people were saying, are game developers now 68 00:02:45,155 --> 00:02:47,974 suddenly gonna decide to optimize their games 69 00:02:48,594 --> 00:02:51,395 for better memory utilization since Crucial is no 70 00:02:51,395 --> 00:02:52,534 longer in the market? 71 00:02:53,129 --> 00:02:55,849 Like, you can't buy machines with massive amounts 72 00:02:55,849 --> 00:02:56,509 of memory. 73 00:02:57,449 --> 00:02:59,549 So we live in this world in processors, 74 00:02:59,610 --> 00:03:02,169 and, Mike, you'll tell us that this is 75 00:03:02,169 --> 00:03:03,310 true of GPUs. 76 00:03:04,169 --> 00:03:07,049 This is true of network processing units, forwarding 77 00:03:07,049 --> 00:03:08,349 engines, right, FEs. 78 00:03:09,705 --> 00:03:10,844 Of all the sidecars 79 00:03:11,145 --> 00:03:13,085 and everything that go with FEs, 80 00:03:14,025 --> 00:03:15,165 including the memory, 81 00:03:16,504 --> 00:03:20,525 the trace motherboards, everything, it's all resource constrained 82 00:03:21,224 --> 00:03:21,965 right now. 83 00:03:23,129 --> 00:03:23,789 And so 84 00:03:24,489 --> 00:03:25,930 there are a lot of side effects of 85 00:03:25,930 --> 00:03:26,430 that, 86 00:03:26,810 --> 00:03:28,969 and they make some differences in the way 87 00:03:28,969 --> 00:03:31,069 you architect things, the way you design things, 88 00:03:31,289 --> 00:03:33,629 the way you even do your career 89 00:03:34,090 --> 00:03:35,709 to some degree or another. 90 00:03:36,275 --> 00:03:37,955 So let's talk through some of those things 91 00:03:37,955 --> 00:03:39,814 because they might be really 92 00:03:40,915 --> 00:03:43,875 very important in the way people think about 93 00:03:43,875 --> 00:03:45,254 what they're working on. 94 00:03:45,875 --> 00:03:47,715 So I don't know. Mike, you wanna start 95 00:03:47,715 --> 00:03:49,895 with whatever's top of top of mind? 96 00:03:50,519 --> 00:03:52,439 Yeah. You mentioned I mean, you you used 97 00:03:52,439 --> 00:03:55,280 the words right now. I think it's the 98 00:03:55,280 --> 00:03:57,479 the the scary part is not that supply 99 00:03:57,479 --> 00:03:58,219 is constrained. 100 00:03:58,919 --> 00:04:00,219 It's what if it's constrained 101 00:04:01,319 --> 00:04:01,819 forever? 102 00:04:04,044 --> 00:04:06,205 Because so here's the thing. During COVID, we 103 00:04:06,205 --> 00:04:07,905 knew that there was a drop in manufacturing 104 00:04:08,125 --> 00:04:10,925 capacity. Right? Literally, people couldn't work, and so 105 00:04:10,925 --> 00:04:12,224 you saw, like, a physical 106 00:04:12,605 --> 00:04:13,905 reduction in 107 00:04:14,365 --> 00:04:16,365 how many, you know, units of whatever could 108 00:04:16,365 --> 00:04:18,205 be produced, you know, in in all these 109 00:04:18,205 --> 00:04:18,959 different plants. 110 00:04:19,680 --> 00:04:21,680 And then we knew that there were changes 111 00:04:21,680 --> 00:04:23,519 in, like, transportation, and so there was a 112 00:04:23,519 --> 00:04:25,439 lot of discussion about, you know, if there's 113 00:04:25,439 --> 00:04:27,839 fewer ships or people can't offload the cargo 114 00:04:27,839 --> 00:04:29,519 at these ports. If you remember all the 115 00:04:29,759 --> 00:04:31,519 like, looking at Los Angeles, the Port Of 116 00:04:31,519 --> 00:04:33,199 Los Angeles and looking at all the the 117 00:04:33,199 --> 00:04:36,785 the ships that were essentially waiting offshore because 118 00:04:36,785 --> 00:04:39,425 they didn't have enough people to to offload 119 00:04:39,425 --> 00:04:39,925 stuff. 120 00:04:40,704 --> 00:04:42,544 And so we we kinda thought that during 121 00:04:42,544 --> 00:04:45,185 COVID, everything was constrained just because the the 122 00:04:45,185 --> 00:04:48,400 conditions led it to that point. What's happened 123 00:04:48,400 --> 00:04:50,560 since COVID is we've never really returned back 124 00:04:50,560 --> 00:04:51,220 to normal. 125 00:04:52,000 --> 00:04:53,600 Like, what if this is the the new 126 00:04:53,600 --> 00:04:55,520 normal? I mean, like, lead times are measured, 127 00:04:55,520 --> 00:04:58,259 you know, typically in in months and quarters. 128 00:04:58,960 --> 00:05:00,480 We're getting close to the point where it's 129 00:05:00,480 --> 00:05:01,540 measured in years, 130 00:05:02,154 --> 00:05:03,455 which is, like, insane. 131 00:05:04,074 --> 00:05:05,995 And then as a result, all the pricing 132 00:05:05,995 --> 00:05:08,555 changes. And so and there's some actually there's 133 00:05:08,555 --> 00:05:10,314 some I know maybe not that interesting to 134 00:05:10,314 --> 00:05:13,035 most people, but, like, there's implications to that. 135 00:05:13,035 --> 00:05:13,615 If I 136 00:05:13,915 --> 00:05:14,895 if I'm spending 137 00:05:15,360 --> 00:05:17,599 if I'm making a purchase decision today, but 138 00:05:17,599 --> 00:05:20,180 it's being shipped next calendar year, 139 00:05:20,560 --> 00:05:22,899 I'm spending next year's money, 140 00:05:23,919 --> 00:05:24,979 next year's budget. 141 00:05:25,599 --> 00:05:27,279 And then if the budget if the cost 142 00:05:27,279 --> 00:05:29,199 of memory as you mentioned, like, it's, like, 143 00:05:29,199 --> 00:05:29,699 doubling, 144 00:05:30,375 --> 00:05:33,095 if the cost of key components is increasing 145 00:05:33,095 --> 00:05:35,654 in that period because everything is kind of 146 00:05:35,654 --> 00:05:36,154 constrained, 147 00:05:37,175 --> 00:05:37,675 then 148 00:05:38,214 --> 00:05:39,595 I can't even use, 149 00:05:40,134 --> 00:05:43,654 like, master purchase agreements to lock in pricing 150 00:05:43,654 --> 00:05:45,175 on that. And so not only am I 151 00:05:45,175 --> 00:05:46,850 spending next year's budget, 152 00:05:47,629 --> 00:05:49,629 I'm not even positive how much of next 153 00:05:49,629 --> 00:05:51,970 year's budget I'm ultimately gonna have to spend. 154 00:05:52,669 --> 00:05:55,250 Yeah. I'm actually giving control of my internal 155 00:05:55,310 --> 00:05:57,250 budget over to a supplier 156 00:05:58,029 --> 00:06:00,589 in a sense. Effectively. And and then and 157 00:06:00,589 --> 00:06:01,089 then 158 00:06:01,524 --> 00:06:02,964 you you can kinda move beyond that, and 159 00:06:02,964 --> 00:06:04,564 you say, well, then, you know, that's that's 160 00:06:04,564 --> 00:06:06,165 even in the best of times. And then 161 00:06:06,165 --> 00:06:08,824 you say, like, during these AI boomed times, 162 00:06:10,324 --> 00:06:11,625 we're gonna spike demand. 163 00:06:12,805 --> 00:06:14,725 And so, like, it's gonna it's gonna make 164 00:06:14,725 --> 00:06:17,449 everything, you know, like like, way, way harder. 165 00:06:17,589 --> 00:06:19,350 Like, that's kind of the environment we're in. 166 00:06:19,350 --> 00:06:20,870 And and so the you know, for me, 167 00:06:20,870 --> 00:06:22,790 the like, I'm I live on the vendor 168 00:06:22,790 --> 00:06:23,529 side. Right? 169 00:06:24,389 --> 00:06:25,910 On the vendor side, you know, you you 170 00:06:25,910 --> 00:06:28,629 talk to customers and they're no longer longer 171 00:06:28,629 --> 00:06:30,225 trying to figure out how do I 172 00:06:30,865 --> 00:06:33,125 connect my servers, how do I 173 00:06:33,425 --> 00:06:35,824 connect my sites. They're figuring out, like, how 174 00:06:35,824 --> 00:06:36,485 do I 175 00:06:37,504 --> 00:06:38,645 financially engineer 176 00:06:39,985 --> 00:06:40,485 the 177 00:06:41,345 --> 00:06:43,285 purchase I have to go make. Or 178 00:06:43,600 --> 00:06:46,639 they're no longer worried about which protocol do 179 00:06:46,639 --> 00:06:48,019 I use. They're like, well, 180 00:06:48,399 --> 00:06:50,639 what does our supply chain resiliency look like? 181 00:06:50,639 --> 00:06:51,379 They're making 182 00:06:51,920 --> 00:06:55,279 architectural decisions based on procurement inputs, which is 183 00:06:55,279 --> 00:06:57,040 something I don't think most people have even 184 00:06:57,040 --> 00:06:57,779 thought about. 185 00:06:58,194 --> 00:06:59,955 Yeah. And so, if you want it like, 186 00:06:59,955 --> 00:07:01,795 as an individual then, if you wanna be 187 00:07:01,795 --> 00:07:04,514 relevant in your company, you gotta realize that 188 00:07:04,514 --> 00:07:06,435 it's more than connecting devices. Now it's like, 189 00:07:06,435 --> 00:07:08,274 how do I navigate all of these different 190 00:07:08,274 --> 00:07:09,955 variables, all these things that are important to 191 00:07:09,955 --> 00:07:11,875 my company? Like, I just think that that 192 00:07:11,875 --> 00:07:13,254 that impact is huge. 193 00:07:13,649 --> 00:07:15,569 And then now when you think about, like, 194 00:07:15,569 --> 00:07:16,389 the architectural 195 00:07:17,490 --> 00:07:17,990 implications 196 00:07:19,329 --> 00:07:21,350 of supply chain based decisions, 197 00:07:21,729 --> 00:07:23,169 like, okay. So now I'm gonna go multi 198 00:07:23,169 --> 00:07:24,769 vendor. Maybe I didn't wanna go multi vendor 199 00:07:24,769 --> 00:07:26,754 before. Now I'm gonna go multi vendor, or 200 00:07:26,754 --> 00:07:28,514 maybe I can't get access to supply, so 201 00:07:28,514 --> 00:07:30,834 I can't go with vendor vendor of choice. 202 00:07:30,834 --> 00:07:32,274 So I'm gonna go, you know, with with 203 00:07:32,274 --> 00:07:34,834 a second vendor, which maybe has like, what 204 00:07:34,834 --> 00:07:36,454 if it's a different operating system? 205 00:07:36,995 --> 00:07:38,274 What if it's like a you know, I 206 00:07:38,274 --> 00:07:40,435 gotta do the there's integration testing, and there's 207 00:07:40,435 --> 00:07:42,509 qualification testing. What if you've got orchestration? 208 00:07:42,889 --> 00:07:45,050 Like, I just the topic is is so 209 00:07:45,050 --> 00:07:46,110 big and hairy. 210 00:07:46,569 --> 00:07:48,250 Like, even just making sense of all of 211 00:07:48,250 --> 00:07:50,649 that. I just it I I don't know 212 00:07:50,649 --> 00:07:53,209 how we we collectively as an industry Yeah. 213 00:07:53,370 --> 00:07:55,544 Make progress. I mean, it it there are 214 00:07:55,544 --> 00:07:57,224 many things you could say come see come 215 00:07:57,224 --> 00:07:59,245 out of this. Like, for instance, 216 00:08:00,024 --> 00:08:02,625 okay, we talked about white box and des 217 00:08:02,664 --> 00:08:03,164 disaggregation 218 00:08:04,185 --> 00:08:06,824 separating our hardware of going to cattle, not 219 00:08:06,824 --> 00:08:07,324 pets. 220 00:08:08,104 --> 00:08:09,004 Oh my goodness. 221 00:08:09,349 --> 00:08:10,729 Cattle not pets. 222 00:08:11,189 --> 00:08:13,910 I mean, making your hardware and your your 223 00:08:13,910 --> 00:08:15,050 your stuff interchangeable 224 00:08:15,589 --> 00:08:17,349 because you never know where you're gonna be 225 00:08:17,349 --> 00:08:18,569 able to get your next 226 00:08:19,029 --> 00:08:21,764 set of 500 routers you might need for 227 00:08:21,764 --> 00:08:24,745 your data center fabric or whatever it is, 228 00:08:25,045 --> 00:08:27,144 or even to do your edge connectivity. 229 00:08:28,485 --> 00:08:28,985 Telemetry 230 00:08:29,605 --> 00:08:32,245 so that that's even right there a starting 231 00:08:32,245 --> 00:08:33,225 point to me 232 00:08:33,684 --> 00:08:34,345 is that 233 00:08:35,179 --> 00:08:37,360 you need to stop thinking as an individual 234 00:08:37,659 --> 00:08:38,159 about, 235 00:08:39,500 --> 00:08:41,899 you know, these are individual devices and I 236 00:08:41,899 --> 00:08:44,639 can do counts or no. It's the hardware 237 00:08:44,700 --> 00:08:45,600 you can get 238 00:08:46,220 --> 00:08:47,679 to get the job done 239 00:08:48,299 --> 00:08:48,799 tomorrow 240 00:08:49,955 --> 00:08:51,095 that you need to do. 241 00:08:51,394 --> 00:08:53,794 Another interesting problem to me there is, what 242 00:08:53,794 --> 00:08:56,514 do I do about financial planning? I've gotta 243 00:08:56,514 --> 00:08:57,654 be able to justify 244 00:08:59,075 --> 00:09:02,034 the cost of buying something two years from 245 00:09:02,034 --> 00:09:04,300 now or three years from now as lead 246 00:09:04,300 --> 00:09:05,840 times might happen to be. 247 00:09:06,300 --> 00:09:07,519 And I don't know 248 00:09:07,980 --> 00:09:09,899 if I'm gonna have the business to do 249 00:09:09,899 --> 00:09:10,399 it. 250 00:09:10,940 --> 00:09:12,860 Well, and it's not it's not and in 251 00:09:12,860 --> 00:09:14,639 in in this constrained environment, 252 00:09:15,100 --> 00:09:16,960 it's not just do I have the money 253 00:09:17,254 --> 00:09:19,334 and the business to do it? Do I 254 00:09:19,334 --> 00:09:20,235 have the scale? 255 00:09:20,695 --> 00:09:22,955 Because in a source constrained environment, 256 00:09:23,575 --> 00:09:26,214 the sellers are going, the people who actually 257 00:09:26,214 --> 00:09:27,195 have the stuff 258 00:09:27,575 --> 00:09:29,894 are going to sell to the people with 259 00:09:29,894 --> 00:09:31,195 the biggest pockets, 260 00:09:31,894 --> 00:09:34,750 deepest pockets who who can who can and 261 00:09:34,750 --> 00:09:36,509 and this is like the way the world 262 00:09:36,509 --> 00:09:38,930 works. I'm not even making a moral judgment 263 00:09:38,990 --> 00:09:41,070 of this. This is just it's what you 264 00:09:41,070 --> 00:09:43,389 would do. Right? If you can only sell 265 00:09:43,389 --> 00:09:44,450 a 100 donuts 266 00:09:44,910 --> 00:09:47,230 and somebody comes along who can buy 99 267 00:09:47,230 --> 00:09:48,764 of them and is willing to pay your 268 00:09:48,924 --> 00:09:51,085 top price, you're going to go ahead and 269 00:09:51,085 --> 00:09:52,384 sell the 99. 270 00:09:52,924 --> 00:09:55,245 And if somebody else comes along and wants 271 00:09:55,245 --> 00:09:57,725 to buy two donuts for them and their 272 00:09:57,725 --> 00:09:59,105 spouse, partner, whatever, 273 00:09:59,565 --> 00:10:00,945 like, you're gonna be like, sorry. 274 00:10:01,404 --> 00:10:04,529 Like, I've promised 99 to this guy. And, 275 00:10:04,529 --> 00:10:07,330 oh, by the way, he can pay 1.52 276 00:10:07,330 --> 00:10:09,990 of what you can pay for 99 277 00:10:10,370 --> 00:10:11,970 even if you're willing to pay five x 278 00:10:11,970 --> 00:10:14,529 for two. It it just it You see 279 00:10:14,529 --> 00:10:16,529 it in a house, right? Math. Like, in 280 00:10:16,529 --> 00:10:18,209 in housing, like, people come in with an 281 00:10:18,209 --> 00:10:20,384 all cash offer. I mean people, you know, 282 00:10:20,384 --> 00:10:21,365 you're you get outbid. 283 00:10:21,985 --> 00:10:23,436 Yeah. It's and it's Right. It's and it's 284 00:10:23,470 --> 00:10:25,184 Right. How frustrating that is, and like and 285 00:10:25,184 --> 00:10:26,945 and you can always go look for another 286 00:10:26,945 --> 00:10:28,784 house. I think when you get into like 287 00:10:28,784 --> 00:10:30,964 IT, it's like it's tougher. Like there's there's 288 00:10:31,264 --> 00:10:32,649 not always another house. You have to, like, 289 00:10:32,649 --> 00:10:34,669 hire a whole different set of people and 290 00:10:35,049 --> 00:10:36,350 figure out how to Something 291 00:10:36,889 --> 00:10:39,529 else I was thinking is you and, as 292 00:10:39,529 --> 00:10:42,269 Mike and Russ were speaking is, you know, 293 00:10:42,330 --> 00:10:45,309 many of us learned to design and architect 294 00:10:45,450 --> 00:10:47,149 and think about infrastructure 295 00:10:48,404 --> 00:10:50,024 inside of a certain set of constraints. 296 00:10:51,125 --> 00:10:54,084 And your mental models may be tuned to 297 00:10:54,084 --> 00:10:54,824 a certain 298 00:10:55,445 --> 00:10:56,424 set of constraints 299 00:10:56,804 --> 00:11:00,184 that you that are unconscious to you. Right? 300 00:11:00,644 --> 00:11:02,745 And so you kind of have to also 301 00:11:02,804 --> 00:11:03,840 peel that back, 302 00:11:04,960 --> 00:11:05,460 reevaluate 303 00:11:05,840 --> 00:11:08,580 what assumptions you may have been making that, 304 00:11:08,720 --> 00:11:09,220 oh, 305 00:11:09,600 --> 00:11:12,559 consistency across the entire environment is the most 306 00:11:12,559 --> 00:11:16,660 important thing. And oh, saving engineer time 307 00:11:17,120 --> 00:11:19,985 is, you know, these things save engineers time 308 00:11:19,985 --> 00:11:22,245 and and you've got a set of architectures 309 00:11:22,464 --> 00:11:26,085 built around that. That may no longer matter 310 00:11:26,945 --> 00:11:27,445 anymore. 311 00:11:27,825 --> 00:11:30,304 And so it's it's important to also acknowledge 312 00:11:30,304 --> 00:11:32,804 like, hey, there's a whole set of assumptions 313 00:11:33,320 --> 00:11:34,860 that may not apply anymore. 314 00:11:35,399 --> 00:11:36,839 Yeah. One of the I feel like one 315 00:11:36,839 --> 00:11:37,959 of the lessons I learned, 316 00:11:38,679 --> 00:11:39,500 from a technical, 317 00:11:40,039 --> 00:11:42,379 subject matter but applies here in other ways. 318 00:11:42,600 --> 00:11:45,080 I remember doing it first learning about vendor 319 00:11:45,080 --> 00:11:46,679 interop like, early in my career, I got 320 00:11:46,679 --> 00:11:48,360 another vendor besides the one I first cut 321 00:11:48,360 --> 00:11:50,694 my teeth on. And, like, not only was 322 00:11:50,694 --> 00:11:52,454 the syntax different, the way that that operating 323 00:11:52,454 --> 00:11:54,855 system thought about hardware was different. And I 324 00:11:54,855 --> 00:11:56,875 remember being just very, like, woah. 325 00:11:57,334 --> 00:11:59,595 Okay. And and maybe do it that way? 326 00:12:00,054 --> 00:12:02,490 Yeah. Yeah. Yeah. Exactly. And it what it 327 00:12:02,490 --> 00:12:03,769 forced me to do is look at the 328 00:12:03,769 --> 00:12:06,009 primitives. Like, look at what's the what the 329 00:12:06,009 --> 00:12:08,570 protocol doing or the more important than the 330 00:12:08,570 --> 00:12:10,410 protocol, like, what's happening on the wire. And 331 00:12:10,410 --> 00:12:12,090 then once I understood that, then adding the 332 00:12:12,090 --> 00:12:14,809 third vendor was really easy, and I felt 333 00:12:14,809 --> 00:12:16,845 really empowered. And I felt a lot more 334 00:12:16,845 --> 00:12:18,945 confident to deal with uncertainty in the future. 335 00:12:19,085 --> 00:12:20,365 And I feel like the what you're talking 336 00:12:20,365 --> 00:12:22,764 about with mental models, Yvonne, that's is is 337 00:12:22,764 --> 00:12:24,764 making me think about that that the you 338 00:12:24,764 --> 00:12:26,625 know, we we have to be able to, 339 00:12:27,965 --> 00:12:30,605 uncertainty cannot can no longer be sort of 340 00:12:30,605 --> 00:12:32,620 this nightmare scenario. We have to it has 341 00:12:32,620 --> 00:12:34,220 a cost attached to it, and it has 342 00:12:34,220 --> 00:12:36,779 constraints attached to it, uncertainty itself. And if 343 00:12:36,779 --> 00:12:38,480 you can face it, I think that that's 344 00:12:38,700 --> 00:12:39,660 a I think it's a really good thing 345 00:12:39,660 --> 00:12:42,160 for individuals and probably for companies as well. 346 00:12:42,540 --> 00:12:44,139 What you just said there is, like, huge 347 00:12:44,139 --> 00:12:45,120 though. So, like, 348 00:12:45,580 --> 00:12:46,320 for you, 349 00:12:46,835 --> 00:12:48,514 you took a thing, and then it you 350 00:12:48,514 --> 00:12:50,274 kinda went deeper and deeper and deeper into 351 00:12:50,274 --> 00:12:52,835 it, which then gave you kinda control. And 352 00:12:52,835 --> 00:12:54,774 so what you ended up with was great. 353 00:12:55,394 --> 00:12:57,394 If you look at that same dynamic happening 354 00:12:57,394 --> 00:12:59,879 now, it's happening at a time where we're 355 00:12:59,879 --> 00:13:01,879 abstracting more and more of the of the 356 00:13:01,879 --> 00:13:04,540 networking detail, the implementation detail away. 357 00:13:05,000 --> 00:13:07,000 We're putting in a fancy GUI, and when, 358 00:13:07,000 --> 00:13:09,559 like, the AI ops sort of rolls through 359 00:13:09,559 --> 00:13:11,960 and everything is a natural language front end, 360 00:13:11,960 --> 00:13:13,320 like, you don't even have the ability to 361 00:13:13,320 --> 00:13:14,379 peel it back anymore. 362 00:13:15,325 --> 00:13:17,004 Like, I that's actually I mean, I didn't 363 00:13:17,565 --> 00:13:19,725 when you said that, I'd I'm it's mildly 364 00:13:19,884 --> 00:13:21,424 I mean, I'm glad I was close to 365 00:13:21,565 --> 00:13:22,845 I'm glad I'm old because then I don't 366 00:13:22,845 --> 00:13:23,745 have to do it. 367 00:13:25,085 --> 00:13:27,485 It is mildly terrifying. But, you know, it's 368 00:13:27,485 --> 00:13:29,245 even more terrifying if you think about it 369 00:13:29,245 --> 00:13:31,200 in terms of what you were saying before 370 00:13:31,200 --> 00:13:33,779 that or what Yvonne was saying, like, okay. 371 00:13:34,159 --> 00:13:35,220 Now the hardware 372 00:13:35,519 --> 00:13:36,240 is not 373 00:13:36,960 --> 00:13:39,039 is is used to be whenever I teach 374 00:13:39,039 --> 00:13:41,519 networking classes. I'm teaching for UC right now. 375 00:13:41,519 --> 00:13:43,394 And the first thing I tell people is, 376 00:13:43,554 --> 00:13:45,475 what is the most expensive thing in the 377 00:13:45,475 --> 00:13:46,855 network? It's the people. 378 00:13:47,475 --> 00:13:50,274 It's the people who sit there and build 379 00:13:50,274 --> 00:13:53,894 the network, manage it, control it, troubleshoot it. 380 00:13:54,115 --> 00:13:56,054 What if that's no longer true? 381 00:13:57,830 --> 00:13:59,289 What if it's the hardware 382 00:14:00,230 --> 00:14:02,649 that becomes the most expensive thing, 383 00:14:02,950 --> 00:14:05,370 the CapEx? What if you simply cannot 384 00:14:05,830 --> 00:14:09,029 replace it? Plant. Right? Like, not even the 385 00:14:09,029 --> 00:14:11,830 hardware, but, like The power. The power and 386 00:14:11,830 --> 00:14:12,330 water. 387 00:14:12,925 --> 00:14:13,985 Yeah. Right. 388 00:14:14,365 --> 00:14:16,445 So what does that do to your entire 389 00:14:16,445 --> 00:14:18,465 concept of life cycle management? 390 00:14:19,325 --> 00:14:21,665 What does that do to your entire concept 391 00:14:22,125 --> 00:14:24,865 of cattle versus pets and your entire concept 392 00:14:25,565 --> 00:14:26,225 of automation 393 00:14:27,019 --> 00:14:29,179 and all of that stuff that goes into 394 00:14:29,179 --> 00:14:29,679 it. 395 00:14:30,860 --> 00:14:32,720 What does that do to vendor differentiation? 396 00:14:33,820 --> 00:14:36,299 Does a vendor now just differentiate themselves because 397 00:14:36,299 --> 00:14:37,679 they have the gear to sell? 398 00:14:38,779 --> 00:14:40,960 I will tell you. Like, that is so, 399 00:14:42,894 --> 00:14:44,575 there there to me, there's there's a couple 400 00:14:44,575 --> 00:14:46,894 of requirements in in networking today. Right? It 401 00:14:46,894 --> 00:14:49,535 just in data center, it just works as, 402 00:14:49,535 --> 00:14:51,695 like, the number one requirement. People just want 403 00:14:51,695 --> 00:14:53,774 stuff that works, and and and we've had 404 00:14:53,774 --> 00:14:56,429 private discussions. Like, that's not always the state 405 00:14:56,429 --> 00:14:58,029 of play today, and so I think that 406 00:14:58,029 --> 00:14:59,569 becomes the thing people care about. 407 00:15:00,110 --> 00:15:02,110 It's easy is is kind of the number 408 00:15:02,110 --> 00:15:03,389 two. Like, I gotta be able to to 409 00:15:03,389 --> 00:15:04,529 kind of work with it. 410 00:15:04,990 --> 00:15:06,350 And so you would think that all the 411 00:15:06,350 --> 00:15:08,774 differentiation would be really around, like, let's say, 412 00:15:08,774 --> 00:15:10,855 hardware and software quality and some of the 413 00:15:10,855 --> 00:15:11,834 operations bits. 414 00:15:12,134 --> 00:15:13,995 The number of deals these days that transact 415 00:15:14,214 --> 00:15:16,454 primarily on, I can ship it in twelve 416 00:15:16,454 --> 00:15:18,394 weeks or in, you know, ten weeks, 417 00:15:18,934 --> 00:15:20,695 like, it is nuts. And so people are 418 00:15:20,774 --> 00:15:22,375 and, like, think about the implication of that. 419 00:15:22,375 --> 00:15:24,169 Imagine you were gonna buy, you know, name 420 00:15:24,169 --> 00:15:26,409 any, you know, vendor you love, and imagine 421 00:15:26,409 --> 00:15:28,730 now that the vendor you don't love has 422 00:15:28,730 --> 00:15:30,889 equipment. And imagine that procurement team then goes 423 00:15:30,889 --> 00:15:32,330 and says they're gonna buy that, but then 424 00:15:32,330 --> 00:15:33,929 they hand that to the team. I mean, 425 00:15:33,929 --> 00:15:35,710 that's, like, that's nuts. 426 00:15:36,595 --> 00:15:38,754 And and and if you push back on 427 00:15:38,754 --> 00:15:40,355 that and be like, well, I don't work 428 00:15:40,355 --> 00:15:41,715 on that. I don't know how to do 429 00:15:41,715 --> 00:15:43,174 that. The answer is, 430 00:15:43,475 --> 00:15:46,035 okay. We're gonna go get somebody who does. 431 00:15:46,035 --> 00:15:47,335 Like, you don't have 432 00:15:47,794 --> 00:15:48,774 leverage there 433 00:15:49,160 --> 00:15:51,320 And you need to, like, understand the world 434 00:15:51,320 --> 00:15:52,220 that you're in 435 00:15:53,000 --> 00:15:55,580 and be, you know, before you 436 00:15:56,240 --> 00:15:56,740 you 437 00:15:57,399 --> 00:16:00,540 you have an overinflated view of your contribution 438 00:16:00,600 --> 00:16:03,019 to the process. We all contribute. It's important. 439 00:16:03,320 --> 00:16:05,184 But you also kind of have to know 440 00:16:05,184 --> 00:16:07,264 what's going on in the world around you. 441 00:16:07,264 --> 00:16:08,865 And and and that's why these kind of 442 00:16:08,865 --> 00:16:10,084 conversations are so important. 443 00:16:10,464 --> 00:16:11,904 Well, that that would I mean, if I 444 00:16:11,904 --> 00:16:14,225 were mentoring somebody and you're in an environment 445 00:16:14,225 --> 00:16:16,464 that's purchasing at enough scale that they're likely 446 00:16:16,464 --> 00:16:19,419 to to go down this path either intentionally 447 00:16:19,419 --> 00:16:21,740 or because they're forced to. Like, now my 448 00:16:21,740 --> 00:16:22,240 mentoring 449 00:16:22,860 --> 00:16:24,559 changes. Right? You know, you should be 450 00:16:25,980 --> 00:16:27,899 or maybe chasing the next cert isn't the 451 00:16:27,899 --> 00:16:29,419 big thing to go do. Like, make sure 452 00:16:29,419 --> 00:16:31,725 you're solid on the underlying bits, and then 453 00:16:31,725 --> 00:16:34,365 start getting experience with virtual labs that give 454 00:16:34,365 --> 00:16:36,924 you access to the different CLIs so that 455 00:16:36,924 --> 00:16:38,764 regardless of which vendor comes down the pike 456 00:16:38,764 --> 00:16:39,264 next, 457 00:16:39,804 --> 00:16:41,565 you know, you're you're not you're not frozen, 458 00:16:41,565 --> 00:16:43,485 you know, during the headlights. But, like, I 459 00:16:43,485 --> 00:16:45,980 think people have to become maybe not multidisciplined, 460 00:16:46,440 --> 00:16:48,059 but but, you know, certainly 461 00:16:49,000 --> 00:16:50,779 broader than, you know, I'm a 462 00:16:51,240 --> 00:16:52,919 a vendor bigots with my you know, this 463 00:16:52,919 --> 00:16:54,279 is, like, the CLI that I know and 464 00:16:54,279 --> 00:16:55,639 love, and I can only work in this 465 00:16:55,639 --> 00:16:56,139 environment. 466 00:16:56,600 --> 00:17:00,024 Yeah. Maybe maybe. And, and Tom and Russell 467 00:17:00,024 --> 00:17:02,664 understand that we we, the last podcast we 468 00:17:02,664 --> 00:17:04,125 recorded was on generalization 469 00:17:04,424 --> 00:17:05,325 versus specialization. 470 00:17:06,265 --> 00:17:07,085 And so 471 00:17:07,384 --> 00:17:09,325 in a time like this, generalization 472 00:17:09,945 --> 00:17:11,164 becomes more important, 473 00:17:12,825 --> 00:17:16,240 across different technologies, different capabilities, different now that 474 00:17:16,240 --> 00:17:18,340 you don't need to be specialized, but, 475 00:17:19,600 --> 00:17:21,859 that breadth can be incredibly important. 476 00:17:22,799 --> 00:17:24,559 Yeah. And if you think about this in 477 00:17:24,559 --> 00:17:25,539 terms of automation 478 00:17:25,920 --> 00:17:26,420 and 479 00:17:27,105 --> 00:17:28,785 that first thing you said, Mike, when you 480 00:17:28,785 --> 00:17:32,144 think about okay. So number one problem, it's 481 00:17:32,144 --> 00:17:35,105 gotta be easy to deploy. Easy it's gotta 482 00:17:35,105 --> 00:17:36,785 work. It's gotta work. It's gotta work in 483 00:17:36,785 --> 00:17:38,785 a way that's easy. I don't wanna think 484 00:17:38,785 --> 00:17:40,144 too hard about doing it. I have other 485 00:17:40,144 --> 00:17:42,289 things I need to think about. But then 486 00:17:42,350 --> 00:17:43,570 I'm resource constrained 487 00:17:43,870 --> 00:17:46,590 or I'm source I'm I'm constrained in what 488 00:17:46,590 --> 00:17:49,309 I can buy. So what that does is 489 00:17:49,309 --> 00:17:50,049 it forces 490 00:17:50,590 --> 00:17:51,970 you to actually build 491 00:17:52,830 --> 00:17:54,450 a a layer of abstraction 492 00:17:55,304 --> 00:17:57,484 between what you're trying to get done 493 00:17:57,865 --> 00:17:59,544 and what you have to get well, the 494 00:17:59,544 --> 00:18:01,484 tools you have to get it done 495 00:18:01,865 --> 00:18:04,845 to get that done. So it almost enforces, 496 00:18:06,424 --> 00:18:09,804 much more of a an intent based 497 00:18:10,210 --> 00:18:11,029 or an automation 498 00:18:11,410 --> 00:18:12,549 first view 499 00:18:12,930 --> 00:18:14,710 that I want to stop caring 500 00:18:15,570 --> 00:18:16,070 about 501 00:18:16,529 --> 00:18:17,430 that CLI. 502 00:18:17,970 --> 00:18:19,650 I want to stop because I don't have 503 00:18:19,650 --> 00:18:21,809 the time, because now I've gotta deal with 504 00:18:21,809 --> 00:18:23,454 five of them. I've gotta be able to 505 00:18:23,454 --> 00:18:25,315 have the flexibility of sourcing 506 00:18:25,615 --> 00:18:26,835 from multiple sources 507 00:18:27,214 --> 00:18:28,575 to get the job done I need to 508 00:18:28,575 --> 00:18:29,315 get done. 509 00:18:29,615 --> 00:18:31,075 So for me to do that, 510 00:18:31,375 --> 00:18:34,434 I need to extract myself from those 511 00:18:35,214 --> 00:18:37,075 hardware specific bits 512 00:18:37,589 --> 00:18:39,269 as it were? What do you think about 513 00:18:39,269 --> 00:18:40,329 that? Is that like 514 00:18:40,630 --> 00:18:41,450 an area? 515 00:18:41,829 --> 00:18:43,690 And and how does that play into optics, 516 00:18:43,750 --> 00:18:46,250 by the way, and and other physical stuff? 517 00:18:46,630 --> 00:18:49,109 You basically just said all SDN really needed 518 00:18:49,109 --> 00:18:51,585 all this time was a pandemic and global 519 00:18:51,585 --> 00:18:52,085 conflict. 520 00:18:53,625 --> 00:18:54,125 I 521 00:18:55,664 --> 00:18:58,085 had a leader once who told me, 522 00:18:58,545 --> 00:19:00,545 we were we were talking about some challenges 523 00:19:00,545 --> 00:19:02,384 and constraints in our business, and one of 524 00:19:02,384 --> 00:19:03,525 the things he said 525 00:19:03,984 --> 00:19:06,005 is constraints breed innovation. 526 00:19:06,839 --> 00:19:08,059 And I think ultimately, 527 00:19:08,759 --> 00:19:10,059 like, this 528 00:19:10,679 --> 00:19:12,220 particular time of constraint 529 00:19:12,920 --> 00:19:15,740 will result in innovation that at some point 530 00:19:16,359 --> 00:19:17,179 will likely 531 00:19:17,640 --> 00:19:20,119 move the point of constraint to another place 532 00:19:20,119 --> 00:19:21,019 in the stack. 533 00:19:22,154 --> 00:19:24,154 But I think that there's a long tail 534 00:19:24,154 --> 00:19:25,994 on this one. Right? And we're not getting 535 00:19:25,994 --> 00:19:28,095 out of this one quickly, I don't believe, 536 00:19:28,634 --> 00:19:31,115 unless there's something on the horizon that I 537 00:19:31,115 --> 00:19:34,130 don't yet have visibility into. But I think 538 00:19:34,130 --> 00:19:35,970 it's gonna it's gonna take a bit for 539 00:19:35,970 --> 00:19:38,130 us to figure out this one. I think 540 00:19:38,130 --> 00:19:38,630 there's 541 00:19:39,409 --> 00:19:41,169 No. It's all good. I I I think 542 00:19:41,169 --> 00:19:43,409 there's an an architectural skill set that can 543 00:19:43,409 --> 00:19:44,609 be helpful if we're if we kind of 544 00:19:44,609 --> 00:19:46,470 narrow it down to talking about network infrastructure 545 00:19:46,529 --> 00:19:47,029 specifically. 546 00:19:47,914 --> 00:19:49,515 Because we talk about if we have to 547 00:19:49,515 --> 00:19:51,755 have multiple suppliers and maybe sometimes we choose 548 00:19:51,755 --> 00:19:53,694 them not even for cost but for availability, 549 00:19:54,234 --> 00:19:55,755 and we have to glue these all together 550 00:19:55,755 --> 00:19:57,835 into something that serves the business. So there's 551 00:19:57,835 --> 00:20:00,255 multiple places to place that abstraction. 552 00:20:00,829 --> 00:20:02,990 So you could but and the architectural decision 553 00:20:02,990 --> 00:20:04,829 to be made is, do we do we 554 00:20:04,910 --> 00:20:07,869 that abstraction above the operating system layer? Like, 555 00:20:07,869 --> 00:20:10,109 do we create orchestration to deal with multiple 556 00:20:10,109 --> 00:20:10,849 quote vendors? 557 00:20:11,549 --> 00:20:13,470 Or do we do we move that that 558 00:20:13,470 --> 00:20:15,954 that point of orchestration down? Do we build 559 00:20:15,954 --> 00:20:17,474 the operating system so that it runs on 560 00:20:17,474 --> 00:20:19,474 all available hardware? And there's trade offs for 561 00:20:19,474 --> 00:20:21,474 both approaches, of course. But but I think 562 00:20:21,474 --> 00:20:23,075 the skill to know where to put the 563 00:20:23,075 --> 00:20:24,754 abstraction and to not just assume that it 564 00:20:24,754 --> 00:20:26,194 always has to live up here or that 565 00:20:26,194 --> 00:20:27,869 it always lives down here, but to be 566 00:20:27,869 --> 00:20:29,789 able to push the abstraction closer and closer 567 00:20:29,789 --> 00:20:31,710 to the metal. And and that I think 568 00:20:31,710 --> 00:20:33,150 that's a I think it's an important thing 569 00:20:33,150 --> 00:20:34,910 for engineers to be able to at least 570 00:20:34,910 --> 00:20:35,650 think about. 571 00:20:36,269 --> 00:20:37,630 You got you got to where I was 572 00:20:37,630 --> 00:20:38,690 going in a much 573 00:20:39,150 --> 00:20:40,394 more intelligent way. 574 00:20:41,835 --> 00:20:43,355 The the whole sonic thing, I mean, like, 575 00:20:43,355 --> 00:20:45,914 it's an interesting space. Right? If if sonic 576 00:20:45,914 --> 00:20:48,494 can unlock so if you're constrained resources 577 00:20:48,875 --> 00:20:50,634 getting, you know, a certain number of chips 578 00:20:50,634 --> 00:20:51,775 from a named supplier, 579 00:20:52,315 --> 00:20:53,295 then having multiple, 580 00:20:53,980 --> 00:20:56,480 you know, conduits into that supply is important. 581 00:20:56,940 --> 00:20:58,539 And if you wanted to if the cost 582 00:20:58,539 --> 00:21:01,019 of that is whatever the operational complexity is, 583 00:21:01,019 --> 00:21:02,619 the question is how do you deal with 584 00:21:02,619 --> 00:21:05,259 that? If you could solve the OS side, 585 00:21:05,259 --> 00:21:07,099 you know, and and Sonic is interesting there. 586 00:21:07,099 --> 00:21:09,075 Because what if Sonic meant that you could 587 00:21:09,075 --> 00:21:11,654 dip into, you know, Edgecore or into, 588 00:21:12,275 --> 00:21:13,095 you know, Celestica 589 00:21:13,634 --> 00:21:15,174 or into, you know, whoever? 590 00:21:15,795 --> 00:21:17,335 That gives people control. 591 00:21:17,954 --> 00:21:19,555 You know, what they pick up in return 592 00:21:19,555 --> 00:21:21,420 is that they have to then manage, they 593 00:21:21,420 --> 00:21:22,900 have to understand how to work with a 594 00:21:22,900 --> 00:21:25,299 sonic distribution and figure out the implications of 595 00:21:25,299 --> 00:21:26,279 life cycle management. 596 00:21:27,539 --> 00:21:28,200 But that's 597 00:21:28,820 --> 00:21:29,299 pushing the 598 00:21:30,100 --> 00:21:31,960 I I like the idea of pushing abstraction, 599 00:21:32,180 --> 00:21:34,279 you know, closer to the metal. That's a 600 00:21:34,884 --> 00:21:36,585 a good way of thinking about it. 601 00:21:36,964 --> 00:21:39,285 And if that if that ultimately checked a 602 00:21:39,285 --> 00:21:41,525 bunch of boxes, then the question is, you 603 00:21:41,525 --> 00:21:44,484 know, what are the downstream architectural implications of 604 00:21:44,484 --> 00:21:46,484 doing that? And I think people have to 605 00:21:46,484 --> 00:21:48,345 start thinking about it. Because right now, like, 606 00:21:48,565 --> 00:21:50,980 if you're into Sonic, right, then it's like 607 00:21:50,980 --> 00:21:53,059 Sonic, you know, represents a certain number of 608 00:21:53,059 --> 00:21:54,679 things. If you're not into Sonic, 609 00:21:55,380 --> 00:21:57,700 it's, you know, it's that OS that cloud 610 00:21:57,700 --> 00:21:58,679 companies use. 611 00:21:59,859 --> 00:22:00,359 Right. 612 00:22:00,740 --> 00:22:01,240 Yep. 613 00:22:02,019 --> 00:22:02,919 Yeah. Exactly. 614 00:22:03,335 --> 00:22:05,015 If you're not big enough to do your 615 00:22:05,015 --> 00:22:07,494 own OS like some companies do. I mean, 616 00:22:07,494 --> 00:22:09,174 there are some companies that don't have open 617 00:22:09,174 --> 00:22:11,255 source. They just run their own OS, and 618 00:22:11,255 --> 00:22:13,174 they run basically bare metal, and that's how 619 00:22:13,174 --> 00:22:15,575 they manage their problem. They just buy white 620 00:22:15,575 --> 00:22:16,890 box, so to speak, 621 00:22:17,450 --> 00:22:19,049 and they just run everything on top of 622 00:22:19,049 --> 00:22:19,789 it. Yeah. 623 00:22:20,170 --> 00:22:22,250 So what are some of the innovations that 624 00:22:22,250 --> 00:22:24,269 could cause a disruption here? 625 00:22:24,890 --> 00:22:27,130 There are a few. First, we could say 626 00:22:27,369 --> 00:22:30,089 I mean, you know, something that can't go 627 00:22:30,089 --> 00:22:32,269 on forever will not go on forever. 628 00:22:33,065 --> 00:22:35,704 Right? That's I mean, that sounds really stupid, 629 00:22:35,704 --> 00:22:37,785 but that's the way And very few things 630 00:22:37,785 --> 00:22:38,684 go on forever. 631 00:22:39,224 --> 00:22:41,384 Yes. Even if they could go on forever, 632 00:22:41,384 --> 00:22:42,365 they generally 633 00:22:42,744 --> 00:22:43,244 don't. 634 00:22:43,785 --> 00:22:44,285 So 635 00:22:44,980 --> 00:22:46,539 let's start here. So we're in a re 636 00:22:46,659 --> 00:22:48,440 we're in a source constrained environment, 637 00:22:48,819 --> 00:22:51,000 particularly FEs, memory, GPUs, 638 00:22:52,659 --> 00:22:54,819 traces on boards, all that kind of stuff 639 00:22:54,819 --> 00:22:56,039 is all very constrained. 640 00:22:57,315 --> 00:22:57,815 So 641 00:22:58,274 --> 00:23:00,434 there's coming a point now where it's gonna 642 00:23:00,434 --> 00:23:02,774 become unaffordable for people to buy 643 00:23:03,154 --> 00:23:04,835 cell phones because the price of cell phones 644 00:23:04,835 --> 00:23:06,694 is going through the roof because of memory. 645 00:23:06,994 --> 00:23:09,315 The price of computers are going through the 646 00:23:09,315 --> 00:23:11,315 roof because of memory. Everything is going through 647 00:23:11,315 --> 00:23:11,730 the roof. 648 00:23:12,289 --> 00:23:14,690 And so this can't go on forever. Right? 649 00:23:14,690 --> 00:23:15,669 There's gotta be 650 00:23:15,970 --> 00:23:16,710 some some way 651 00:23:17,250 --> 00:23:20,210 that this gets resolved or something happens to 652 00:23:20,210 --> 00:23:20,710 change. 653 00:23:21,250 --> 00:23:22,929 So what are some of those things? We 654 00:23:22,929 --> 00:23:24,230 could have a software breakthrough. 655 00:23:24,744 --> 00:23:27,144 We could have, oh, all of the stuff 656 00:23:27,144 --> 00:23:28,924 we need for AI today 657 00:23:29,944 --> 00:23:31,804 can be solved in software 658 00:23:32,585 --> 00:23:34,204 in a quarter of the time 659 00:23:35,224 --> 00:23:36,924 because some new software 660 00:23:38,090 --> 00:23:39,390 thing that comes out. 661 00:23:40,170 --> 00:23:42,170 What was it I saw? There was a 662 00:23:42,170 --> 00:23:42,670 Chinese 663 00:23:43,210 --> 00:23:45,529 large language model that took up a tenth 664 00:23:45,529 --> 00:23:47,470 of the I don't know where that went, 665 00:23:47,609 --> 00:23:49,710 but it took up a tenth of the 666 00:23:49,769 --> 00:23:51,974 processing power or something like that. 667 00:23:53,015 --> 00:23:55,974 Yeah. Like an algorithmic innovation. Right? Like something 668 00:23:55,974 --> 00:23:59,515 that that dramatically reduces the processing power required 669 00:24:00,134 --> 00:24:02,775 so people just need less. Yep. So that's 670 00:24:02,775 --> 00:24:03,835 that's one possibility. 671 00:24:04,519 --> 00:24:07,000 And that seems like it's not remote. It 672 00:24:07,000 --> 00:24:08,700 seems like it's quite possible 673 00:24:09,159 --> 00:24:10,539 in the history of software. 674 00:24:11,240 --> 00:24:12,619 Another one would be 675 00:24:13,559 --> 00:24:16,379 something like CPO, co packaged optics, 676 00:24:17,964 --> 00:24:19,424 causing a shift 677 00:24:20,125 --> 00:24:22,144 in the way we process packets 678 00:24:23,164 --> 00:24:25,585 such that processing packets becomes 679 00:24:26,044 --> 00:24:28,224 much less expensive or much more efficient. 680 00:24:29,244 --> 00:24:31,289 Any thoughts on that, Mike? Am I in 681 00:24:31,289 --> 00:24:33,529 the wrong area, wrong ocean there? Is that 682 00:24:33,529 --> 00:24:35,450 something? I mean, see the the tests out 683 00:24:35,450 --> 00:24:37,950 of meta on CPO on on both reliability. 684 00:24:38,090 --> 00:24:40,430 The the fear with CPO is that you 685 00:24:40,650 --> 00:24:42,410 if you then have an optic failure, which 686 00:24:42,410 --> 00:24:44,170 is, you know, historically the thing that went 687 00:24:44,170 --> 00:24:46,184 down the most, You have to replace the 688 00:24:46,184 --> 00:24:48,664 entire box. You know, Meta's reports out of 689 00:24:48,664 --> 00:24:51,144 OCP last year saying that they were getting, 690 00:24:51,144 --> 00:24:53,085 you know, fewer failures 691 00:24:53,944 --> 00:24:56,505 overall and and and lower downtime than they 692 00:24:56,505 --> 00:24:58,765 were seeing with, like, their their mainline network. 693 00:24:59,250 --> 00:25:01,329 Like, that's yeah. I mean, CPO is gonna 694 00:25:01,329 --> 00:25:02,950 gonna going to have some effect 695 00:25:03,250 --> 00:25:05,089 in the scale up arena, like, on the 696 00:25:05,089 --> 00:25:07,329 AI side. You're gonna see CPO and scale 697 00:25:07,329 --> 00:25:09,650 up come together. And so what if that 698 00:25:09,650 --> 00:25:11,170 comes together and then you see, you know, 699 00:25:11,170 --> 00:25:13,190 dramatic improvements in in power? 700 00:25:13,515 --> 00:25:15,434 You know, maybe that increases the the, you 701 00:25:15,434 --> 00:25:16,174 know, the density 702 00:25:17,115 --> 00:25:18,015 and that 703 00:25:18,315 --> 00:25:20,394 which, you know, and and and drives cost 704 00:25:20,394 --> 00:25:22,394 down, and and so it it it changes, 705 00:25:22,394 --> 00:25:24,075 you know, where you can build. There's a 706 00:25:24,075 --> 00:25:25,855 lot of implications of that, and 707 00:25:26,154 --> 00:25:27,450 and I think if you if you do 708 00:25:27,450 --> 00:25:29,609 that, then the economics change and and all 709 00:25:29,609 --> 00:25:30,490 of a sudden the, 710 00:25:31,450 --> 00:25:32,970 the number of units that you need to 711 00:25:32,970 --> 00:25:35,450 to purchase to satisfy a workload goes down. 712 00:25:35,450 --> 00:25:37,289 And if that's the case, then, you know, 713 00:25:37,289 --> 00:25:39,529 maybe some of the the constraints go away 714 00:25:39,529 --> 00:25:41,295 and then it opens things up. Like, that's 715 00:25:41,295 --> 00:25:42,195 that's a possibility. 716 00:25:42,894 --> 00:25:44,815 Yeah. There's a bunch of investments and that's 717 00:25:44,815 --> 00:25:48,174 in in similar sounding things. I mentioned we 718 00:25:48,174 --> 00:25:50,174 were chatting earlier, but there's a whole movement 719 00:25:50,174 --> 00:25:52,095 around COOS, you know, chip on wafer on 720 00:25:52,095 --> 00:25:52,595 substrate. 721 00:25:52,975 --> 00:25:54,894 You know, imagine if the packaging wasn't like 722 00:25:54,894 --> 00:25:56,829 on, you know, the the chip or the 723 00:25:56,829 --> 00:25:58,589 dye. What if the packaging was bigger? You 724 00:25:58,589 --> 00:26:00,130 put entire systems together. 725 00:26:00,509 --> 00:26:02,669 You can condense things into very, very small 726 00:26:02,669 --> 00:26:05,009 form factors, and then the the cooling properties 727 00:26:05,069 --> 00:26:05,809 change dramatically. 728 00:26:06,990 --> 00:26:08,349 You know, I mentioned I was talking to 729 00:26:08,349 --> 00:26:10,634 a company that effectively says they think they 730 00:26:10,634 --> 00:26:12,634 can reduce a a whole row of equipment 731 00:26:12,634 --> 00:26:14,315 to something the size of a coffee table. 732 00:26:14,315 --> 00:26:15,695 Like, what if if that happened? 733 00:26:16,075 --> 00:26:17,835 Like, what would the implications be? I mean, 734 00:26:17,835 --> 00:26:20,075 there's, like, it's it's hard to imagine. Like, 735 00:26:20,075 --> 00:26:21,755 that kind of density, you know, it it 736 00:26:21,755 --> 00:26:22,815 it changes everything. 737 00:26:23,740 --> 00:26:25,500 Something like that would would change the way 738 00:26:25,500 --> 00:26:27,279 we think about infrastructure entirely. 739 00:26:28,140 --> 00:26:30,619 And then your constrained resources become, you know, 740 00:26:30,619 --> 00:26:32,079 the it it it 741 00:26:32,380 --> 00:26:33,660 I think, you know, Yvonne, you said that, 742 00:26:33,660 --> 00:26:35,339 you know, the bottleneck moved to somewhere else. 743 00:26:35,339 --> 00:26:36,619 I mean, that would that would move the 744 00:26:36,619 --> 00:26:37,119 bottleneck. 745 00:26:37,795 --> 00:26:40,035 Yeah. Yeah. I know that Jeff Tensorro said 746 00:26:40,035 --> 00:26:41,555 the other day on a podcast I was 747 00:26:41,555 --> 00:26:42,375 on with him 748 00:26:42,755 --> 00:26:43,255 that 749 00:26:43,795 --> 00:26:45,414 NVIDIA, I think it is, 750 00:26:45,875 --> 00:26:46,375 has 751 00:26:47,075 --> 00:26:50,214 CPO co packaged optics that are field replaceable 752 00:26:50,275 --> 00:26:52,115 now, and that may be involved with the 753 00:26:52,115 --> 00:26:52,615 meta 754 00:26:53,230 --> 00:26:54,690 work that you're talking about. 755 00:26:54,990 --> 00:26:56,690 Now on the other side, there are naysayers 756 00:26:56,750 --> 00:26:57,329 for that. 757 00:26:57,950 --> 00:27:00,029 When I was at AutoCon, I ran into 758 00:27:00,029 --> 00:27:01,009 the most interesting 759 00:27:01,389 --> 00:27:01,889 person, 760 00:27:02,349 --> 00:27:05,389 weird that this happened. I ran into Lynn 761 00:27:05,389 --> 00:27:05,889 Bozak. 762 00:27:07,974 --> 00:27:09,835 Let's go. Yeah. Who was there, 763 00:27:10,134 --> 00:27:12,234 and he was there representing, 764 00:27:13,414 --> 00:27:15,174 what's the name of his company? Let me 765 00:27:15,174 --> 00:27:15,994 see. It's, 766 00:27:16,535 --> 00:27:17,035 XKL, 767 00:27:17,654 --> 00:27:19,920 which is a an optics company, and they're 768 00:27:19,920 --> 00:27:20,460 mostly focused 769 00:27:20,920 --> 00:27:23,660 on long haul. And he was telling me 770 00:27:23,880 --> 00:27:25,640 he's pretty convinced that you can't get the 771 00:27:25,640 --> 00:27:27,340 power out of c p CPO 772 00:27:27,720 --> 00:27:30,140 co pass co packaged optics that you need 773 00:27:30,600 --> 00:27:33,180 for much more than, like, short runs between 774 00:27:33,240 --> 00:27:33,674 racks. 775 00:27:34,475 --> 00:27:36,735 So I don't know. You know, there's naysayers 776 00:27:36,795 --> 00:27:39,455 in that space. Well, but even your naysayer 777 00:27:39,515 --> 00:27:41,355 opens up an option. Right? So let's say 778 00:27:41,355 --> 00:27:42,795 that that he's right. Let's say that all 779 00:27:42,795 --> 00:27:44,394 the power goes the other way. Let's say 780 00:27:44,394 --> 00:27:46,015 that instead of seeing vast improvements, 781 00:27:46,759 --> 00:27:48,279 we see that the power you know, right 782 00:27:48,279 --> 00:27:50,600 now we're going a 120 kilowatts per rack 783 00:27:50,600 --> 00:27:52,680 to 600 kilowatts per rack with the Rubin 784 00:27:52,680 --> 00:27:53,180 stuff. 785 00:27:53,799 --> 00:27:55,799 What if what if we the the what 786 00:27:55,799 --> 00:27:57,259 if the infrastructure outpaces 787 00:27:58,039 --> 00:27:59,980 the the power generation? Then 788 00:28:00,554 --> 00:28:02,875 maybe the innovation so if innovation if constraints 789 00:28:02,875 --> 00:28:05,674 create innovation, then maybe we start doing more 790 00:28:05,674 --> 00:28:07,294 work over wide area. 791 00:28:07,595 --> 00:28:09,914 And so instead of having big centralized data 792 00:28:09,914 --> 00:28:12,234 centers, you have federated, like, very distributed data 793 00:28:12,234 --> 00:28:14,289 centers. Yeah. In Europe, I I don't know 794 00:28:14,289 --> 00:28:15,730 if I if they're I I've I've read 795 00:28:15,730 --> 00:28:16,849 it, so I don't know how to pronounce 796 00:28:16,849 --> 00:28:17,990 it. Is it SYNNQ? 797 00:28:18,369 --> 00:28:20,289 S y n n q? Like, they're they're 798 00:28:20,450 --> 00:28:23,089 they've set up a federation between Finland. They're 799 00:28:23,089 --> 00:28:24,309 doing AI clusters 800 00:28:24,769 --> 00:28:26,869 between Finland and and and Italy. 801 00:28:27,704 --> 00:28:29,784 And so they're putting geographical distance. And if 802 00:28:29,784 --> 00:28:32,105 that happened, like, what if the future of 803 00:28:32,105 --> 00:28:32,924 it wasn't 804 00:28:33,384 --> 00:28:35,704 super density? It was like super distribution. Like, 805 00:28:35,704 --> 00:28:37,244 that would also change everything. 806 00:28:37,625 --> 00:28:39,865 And and that would have interesting implications on 807 00:28:39,865 --> 00:28:41,244 things like data sovereignty 808 00:28:42,680 --> 00:28:45,240 and those kinds of topics. Right? Yeah. So 809 00:28:45,320 --> 00:28:47,180 and and so that is a new, 810 00:28:47,960 --> 00:28:50,200 I you know, they there could be advantages. 811 00:28:50,200 --> 00:28:52,440 There could be new constraints there. Right? 812 00:28:53,079 --> 00:28:53,579 So, 813 00:28:53,880 --> 00:28:56,380 yes. I I I'm I'm confident that 814 00:28:58,544 --> 00:29:01,204 the the the market will find a solution, 815 00:29:01,345 --> 00:29:03,904 and eventually, we'll see a new area of 816 00:29:03,904 --> 00:29:04,404 constraint. 817 00:29:04,704 --> 00:29:07,365 Right? The the the challenge is 818 00:29:08,865 --> 00:29:12,029 either either predicting or then navigating in these 819 00:29:12,109 --> 00:29:13,009 in these times 820 00:29:13,630 --> 00:29:17,009 of severe severe constraint, frankly. Because if you're 821 00:29:17,390 --> 00:29:19,330 if if you're a smaller purchaser 822 00:29:19,869 --> 00:29:23,170 right now and you desperately need certain gear, 823 00:29:23,390 --> 00:29:24,930 it's really, really challenging, 824 00:29:25,575 --> 00:29:27,434 which which is gonna force you to innovate. 825 00:29:28,615 --> 00:29:29,115 Yeah. 826 00:29:30,054 --> 00:29:30,554 Yeah. 827 00:29:30,934 --> 00:29:32,214 Well, so what do you do? I mean, 828 00:29:32,214 --> 00:29:34,214 like so if you're let's say you're, like, 829 00:29:34,214 --> 00:29:36,294 a medium sized purchaser. If you're a small 830 00:29:36,294 --> 00:29:38,075 purchaser, you might lack the 831 00:29:38,730 --> 00:29:40,509 the human power and the, 832 00:29:40,970 --> 00:29:42,730 you know, and then the purchasing power to 833 00:29:42,730 --> 00:29:43,929 to do much about it. But if you're, 834 00:29:43,929 --> 00:29:44,890 let's say, you're in the middle of it. 835 00:29:44,890 --> 00:29:46,490 You're not at the at the hyper scale 836 00:29:46,490 --> 00:29:48,170 level of consumption, but you're not at, you 837 00:29:48,170 --> 00:29:49,950 know, Joe's Tire Shop level 838 00:29:50,250 --> 00:29:51,230 of of of consumption. 839 00:29:51,904 --> 00:29:53,184 How do you even, like, how do you 840 00:29:53,184 --> 00:29:54,164 prepare architecturally, 841 00:29:55,664 --> 00:29:56,164 operationally, 842 00:29:56,944 --> 00:29:58,944 even, like, from a workforce perspective? How do 843 00:29:58,944 --> 00:29:59,605 you prepare? 844 00:30:00,544 --> 00:30:02,464 Because if you don't my my, I guess, 845 00:30:02,464 --> 00:30:04,884 my my hesitance is that 846 00:30:05,419 --> 00:30:06,319 if you go 847 00:30:06,779 --> 00:30:07,279 in, 848 00:30:08,539 --> 00:30:09,500 and and you don't know 849 00:30:10,140 --> 00:30:11,500 you haven't thought it through and you put 850 00:30:11,500 --> 00:30:13,819 nothing in place, then you're gonna find out 851 00:30:13,819 --> 00:30:15,980 that, like, the people are gonna you're gonna 852 00:30:15,980 --> 00:30:17,839 rebel, and it's like you can't do anything. 853 00:30:17,900 --> 00:30:18,345 Or 854 00:30:18,744 --> 00:30:21,544 your tooling isn't going to extend to the 855 00:30:21,544 --> 00:30:22,684 new thing. And so 856 00:30:23,065 --> 00:30:24,984 you're architecturally kinda locked in, and you don't 857 00:30:24,984 --> 00:30:26,904 have the ability to even give yourself choice 858 00:30:26,904 --> 00:30:29,304 and control. I mean, how do you how 859 00:30:29,304 --> 00:30:31,470 do you even prepare to get control back? 860 00:30:31,549 --> 00:30:32,929 I think you have to think about 861 00:30:33,230 --> 00:30:34,769 I think you have to think about portability. 862 00:30:34,990 --> 00:30:36,829 Like, as a Yeah. A lot of us 863 00:30:36,829 --> 00:30:39,230 have we have these mantras when we're building 864 00:30:39,230 --> 00:30:40,210 something. Survivability 865 00:30:40,509 --> 00:30:41,970 is really important or, 866 00:30:42,750 --> 00:30:46,029 you know, or resiliency is really resilience is 867 00:30:46,029 --> 00:30:48,164 really the biggest thing we've gotta deal with 868 00:30:48,164 --> 00:30:49,765 and everything else could takes a second chair 869 00:30:49,765 --> 00:30:50,345 to that. 870 00:30:50,724 --> 00:30:52,804 I think one of the having as a 871 00:30:52,804 --> 00:30:54,345 design mantra that portability 872 00:30:54,644 --> 00:30:56,724 is is the most important building this thing 873 00:30:56,724 --> 00:30:58,804 to be moved to another thing, and we 874 00:30:58,804 --> 00:31:00,984 will sacrifice resilience and we will sacrifice 875 00:31:01,650 --> 00:31:02,150 survivability 876 00:31:02,450 --> 00:31:03,190 for this. 877 00:31:04,049 --> 00:31:04,789 And optimization. 878 00:31:05,809 --> 00:31:06,869 Yeah. And optimization. 879 00:31:07,250 --> 00:31:07,750 Yes. 880 00:31:08,210 --> 00:31:08,710 Exactly. 881 00:31:09,170 --> 00:31:10,450 Should be one of the first things to 882 00:31:10,450 --> 00:31:11,650 go in my opinion and one of the 883 00:31:11,650 --> 00:31:13,410 last things that engineers ever wanna let go 884 00:31:13,410 --> 00:31:14,965 of. Right? Yeah. Exactly. 885 00:31:15,424 --> 00:31:15,924 Yeah. 886 00:31:16,225 --> 00:31:17,664 So I would start I would start with 887 00:31:17,664 --> 00:31:19,924 the tooling. Right? That's where I would start. 888 00:31:20,065 --> 00:31:22,065 I mean, start with something that's an open 889 00:31:22,065 --> 00:31:22,565 source, 890 00:31:23,424 --> 00:31:24,484 widely available 891 00:31:25,424 --> 00:31:27,285 tooling system that orchestrates 892 00:31:27,664 --> 00:31:28,884 many different platforms, 893 00:31:29,670 --> 00:31:31,750 extend it to the platforms that you need 894 00:31:31,750 --> 00:31:32,890 to extend it to, 895 00:31:33,190 --> 00:31:35,430 find people to extend it to those platforms 896 00:31:35,430 --> 00:31:36,809 as the case might be. 897 00:31:37,430 --> 00:31:39,110 Pay people to do it if you need 898 00:31:39,110 --> 00:31:41,269 to. They go back to Yvonne's. There are 899 00:31:41,269 --> 00:31:43,690 contractors out there. You can go hire somebody. 900 00:31:44,144 --> 00:31:46,805 Oh, this orchestration system I want to use 901 00:31:46,865 --> 00:31:47,924 doesn't orchestrate, 902 00:31:48,785 --> 00:31:49,605 I don't know, 903 00:31:50,065 --> 00:31:52,865 vendor x. Fine. Go find somebody who can 904 00:31:52,865 --> 00:31:54,945 write that code and get it done so 905 00:31:54,945 --> 00:31:56,565 you can orchestrate to that 906 00:31:57,079 --> 00:31:59,900 and start building an abstraction there between 907 00:32:00,440 --> 00:32:01,740 you and the hardware, 908 00:32:02,200 --> 00:32:03,339 between your engineering 909 00:32:03,640 --> 00:32:04,539 and the hardware. 910 00:32:05,079 --> 00:32:06,859 Perhaps to me is like 911 00:32:07,320 --> 00:32:09,339 a key way to build that portability 912 00:32:09,640 --> 00:32:10,140 Tom's 913 00:32:10,680 --> 00:32:13,234 thinking about or talking about and, 914 00:32:13,934 --> 00:32:15,775 you know, and and be willing to give 915 00:32:15,775 --> 00:32:16,275 up 916 00:32:16,734 --> 00:32:18,835 the optimization, some of the optimization 917 00:32:19,375 --> 00:32:20,275 to get there. 918 00:32:21,215 --> 00:32:22,115 But I think 919 00:32:22,974 --> 00:32:23,474 yeah. 920 00:32:23,775 --> 00:32:25,215 Where I was gonna go in in a 921 00:32:25,215 --> 00:32:27,650 much more generic kind of a way is 922 00:32:27,650 --> 00:32:29,490 you got to figure out what you've really 923 00:32:29,490 --> 00:32:30,309 got to have. 924 00:32:30,769 --> 00:32:32,210 What do we really have to have and 925 00:32:32,210 --> 00:32:33,750 what are the things that we can sacrifice? 926 00:32:33,809 --> 00:32:35,329 And that's, that's kind of what Tom was 927 00:32:35,329 --> 00:32:37,029 talking about by saying maybe, 928 00:32:37,650 --> 00:32:39,990 maybe resilience isn't our number one 929 00:32:40,484 --> 00:32:41,944 design goal now, 930 00:32:42,244 --> 00:32:43,924 right? These are the things that we have 931 00:32:43,924 --> 00:32:46,244 to have to deliver and to keep the 932 00:32:46,244 --> 00:32:49,525 business running and everything else becomes secondary to 933 00:32:49,525 --> 00:32:51,704 that. And that can be really challenging 934 00:32:52,325 --> 00:32:53,384 when you've got, 935 00:32:53,750 --> 00:32:55,990 when you've been operating under a in a 936 00:32:55,990 --> 00:32:57,830 different environment for a long, long time. But 937 00:32:57,830 --> 00:32:58,630 I mean, ultimately 938 00:32:58,950 --> 00:33:00,309 and then I think you have to talk 939 00:33:00,309 --> 00:33:00,809 about 940 00:33:01,269 --> 00:33:03,109 what you're doing and why. Like these are 941 00:33:03,269 --> 00:33:05,029 this is what we have to do. This 942 00:33:05,029 --> 00:33:06,894 is what we'd like to do. These are 943 00:33:06,894 --> 00:33:09,295 the constraints that we're facing. There's this great 944 00:33:09,295 --> 00:33:11,615 moment in the movie Apollo 13, right, where 945 00:33:11,615 --> 00:33:13,934 they realize that there's been an explosion on 946 00:33:13,934 --> 00:33:15,454 the lunar module and they've got to get 947 00:33:15,454 --> 00:33:17,634 people home and they can't filter the air 948 00:33:17,934 --> 00:33:20,414 and and they're the interfaces don't match and 949 00:33:20,414 --> 00:33:22,130 somebody walks in the room with a bag 950 00:33:22,130 --> 00:33:23,809 and dumps it out on the table and 951 00:33:23,809 --> 00:33:24,309 says, 952 00:33:24,690 --> 00:33:26,849 this is what we have to keep them 953 00:33:26,849 --> 00:33:28,609 alive and get them home. Now go figure 954 00:33:28,609 --> 00:33:30,049 it out. I mean, you've got to kind 955 00:33:30,049 --> 00:33:32,450 of have that moment, right? This is what 956 00:33:32,450 --> 00:33:34,369 we have to keep our business alive and 957 00:33:34,369 --> 00:33:36,015 these are what things we need to work. 958 00:33:36,335 --> 00:33:37,855 Now. We've got to figure it out with 959 00:33:37,855 --> 00:33:39,775 these things. Yeah. Well, you have to move 960 00:33:39,775 --> 00:33:42,414 your optimization other places sometimes. Right? And your 961 00:33:42,414 --> 00:33:44,414 resilience other places. You can't count on it 962 00:33:44,414 --> 00:33:46,815 all the time to be there. Like when 963 00:33:46,815 --> 00:33:48,654 I teach data center fabrics, I always talk 964 00:33:48,654 --> 00:33:49,154 about 965 00:33:49,470 --> 00:33:50,210 the importance 966 00:33:50,910 --> 00:33:51,730 of minimizing 967 00:33:52,029 --> 00:33:54,509 jitter. Well, you know what? You may have 968 00:33:54,509 --> 00:33:55,890 to rejigger your application 969 00:33:56,910 --> 00:33:59,470 to support a data center fabric with higher 970 00:33:59,470 --> 00:33:59,970 jitter 971 00:34:00,269 --> 00:34:02,929 because you just can't buy the hardware 972 00:34:04,105 --> 00:34:05,884 that you need to get to 973 00:34:06,424 --> 00:34:08,505 to build the application the way you want 974 00:34:08,505 --> 00:34:10,284 to. That's a good example. 975 00:34:11,065 --> 00:34:12,204 It just may be 976 00:34:12,905 --> 00:34:13,565 a a thing. 977 00:34:14,025 --> 00:34:15,864 Oh, well, that might mean I need to 978 00:34:15,864 --> 00:34:18,859 distribute my front end, as Mike says distributed 979 00:34:19,000 --> 00:34:20,619 computing, more distributed stuff, 980 00:34:21,639 --> 00:34:23,960 out to a CDN instead of always storing 981 00:34:23,960 --> 00:34:25,099 my front end server 982 00:34:25,480 --> 00:34:27,960 in my own data center hosting in my 983 00:34:27,960 --> 00:34:28,859 own data center. 984 00:34:29,239 --> 00:34:29,739 Okay? 985 00:34:30,664 --> 00:34:32,204 Well Maybe. Yeah. 986 00:34:32,744 --> 00:34:33,405 You know? 987 00:34:34,984 --> 00:34:37,324 That's that's the decision you've gotta make. 988 00:34:38,344 --> 00:34:39,085 And so 989 00:34:39,625 --> 00:34:41,625 yeah. And that goes back to Tom's point 990 00:34:41,625 --> 00:34:42,284 on portability 991 00:34:42,744 --> 00:34:43,405 right there. 992 00:34:44,264 --> 00:34:45,885 Build things so they can be 993 00:34:46,219 --> 00:34:47,519 moved other places. 994 00:34:48,460 --> 00:34:50,299 I think some of the text coming along 995 00:34:50,299 --> 00:34:51,659 to make it to make some of the 996 00:34:51,659 --> 00:34:52,719 abstraction possible. 997 00:34:53,659 --> 00:34:54,139 I'm not 998 00:34:55,260 --> 00:34:56,799 I tend to be more of an AI 999 00:34:57,019 --> 00:34:58,940 pragmatist. I don't know that I'd like to 1000 00:34:58,940 --> 00:35:01,324 breathe a lot of additional hype into that 1001 00:35:01,324 --> 00:35:02,144 bubble, but, 1002 00:35:02,925 --> 00:35:04,525 but I I do like natural language as 1003 00:35:04,525 --> 00:35:06,045 a front end as a means of providing 1004 00:35:06,045 --> 00:35:06,545 abstraction. 1005 00:35:07,405 --> 00:35:08,605 Because I think if you can get through 1006 00:35:08,605 --> 00:35:10,204 a lot of the day to day using 1007 00:35:10,204 --> 00:35:11,585 natural language, then 1008 00:35:12,045 --> 00:35:14,460 some of the reliance on on, you know, 1009 00:35:14,460 --> 00:35:16,940 rote memorization of complex syntax. If if you 1010 00:35:16,940 --> 00:35:18,400 can get rid of that a little bit, 1011 00:35:18,539 --> 00:35:20,380 then I think your workforce can kinda come 1012 00:35:20,380 --> 00:35:22,059 along. I do think it comes at the 1013 00:35:22,059 --> 00:35:23,900 expense of knowing the details, which I think, 1014 00:35:23,900 --> 00:35:25,359 you know, that part is 1015 00:35:25,659 --> 00:35:27,614 that's a that's a difficult trade off to 1016 00:35:27,614 --> 00:35:28,914 make in in the absolution. 1017 00:35:29,215 --> 00:35:31,454 I think in in different parts of your 1018 00:35:31,454 --> 00:35:32,994 workforce, it's a little bit easier. 1019 00:35:34,414 --> 00:35:35,454 But I think that gives you at least 1020 00:35:35,454 --> 00:35:37,074 buy you a little bit of breathing room. 1021 00:35:37,934 --> 00:35:39,855 I I'd be shocked if the if the 1022 00:35:39,855 --> 00:35:42,014 interface of the future is not closer to 1023 00:35:42,014 --> 00:35:42,914 natural language. 1024 00:35:44,570 --> 00:35:45,929 You know, and then and then use the 1025 00:35:45,929 --> 00:35:48,570 complex queries for troubleshooting and and monitoring that 1026 00:35:48,570 --> 00:35:50,090 kind of stuff. Like, make that stuff, you 1027 00:35:50,090 --> 00:35:51,050 know, quite a bit more, 1028 00:35:52,010 --> 00:35:53,469 specific and and prescriptive. 1029 00:35:54,329 --> 00:35:54,829 Yeah. 1030 00:35:55,210 --> 00:35:56,969 Yeah. And and and the thing with natural 1031 00:35:56,969 --> 00:35:59,364 language and AI front ends is what always 1032 00:35:59,364 --> 00:36:01,065 scares me a bit is that 1033 00:36:02,164 --> 00:36:03,864 you have a transition period 1034 00:36:04,485 --> 00:36:06,825 where there are people who lived 1035 00:36:08,005 --> 00:36:10,805 growing up and they and they knew the 1036 00:36:10,805 --> 00:36:11,305 technology. 1037 00:36:12,005 --> 00:36:12,505 Right? 1038 00:36:12,840 --> 00:36:15,079 And so the AI LLM or the natural 1039 00:36:15,079 --> 00:36:15,980 language interface 1040 00:36:16,680 --> 00:36:19,579 actually helps them to create better and faster. 1041 00:36:20,360 --> 00:36:22,360 And then you have a a generation that 1042 00:36:22,360 --> 00:36:23,579 comes after them 1043 00:36:24,200 --> 00:36:26,140 that doesn't know anything but 1044 00:36:26,735 --> 00:36:27,235 that 1045 00:36:27,695 --> 00:36:29,235 natural language interface. 1046 00:36:29,695 --> 00:36:30,914 And now the productivity 1047 00:36:31,215 --> 00:36:32,355 goes through the floor 1048 00:36:33,135 --> 00:36:35,934 and everything becomes stale. Okay. And I'm gonna 1049 00:36:35,934 --> 00:36:37,855 give a really weird example of this in 1050 00:36:37,855 --> 00:36:39,454 my head. But one of the ways I 1051 00:36:39,454 --> 00:36:41,315 think about it is in music. 1052 00:36:41,660 --> 00:36:43,340 So I know on x and various other 1053 00:36:43,340 --> 00:36:45,360 places, I hear people say all the time, 1054 00:36:45,420 --> 00:36:46,480 even my kids 1055 00:36:47,099 --> 00:36:49,280 who are not that young any longer, 1056 00:36:49,579 --> 00:36:51,760 but they still listen to the same music 1057 00:36:51,900 --> 00:36:53,440 I listened to as a teenager. 1058 00:36:55,019 --> 00:36:56,640 Right? And you think, 1059 00:36:57,065 --> 00:36:59,405 that's a twenty something year difference. 1060 00:37:00,345 --> 00:37:02,845 Why are they listening to that music? 1061 00:37:03,224 --> 00:37:04,664 And to me, if I look at the 1062 00:37:04,664 --> 00:37:05,644 music industry, 1063 00:37:06,025 --> 00:37:08,445 we started out with marambas and 1064 00:37:09,239 --> 00:37:12,440 acoustic guitars, and then they added electronics to 1065 00:37:12,440 --> 00:37:14,300 it. And there was a point 1066 00:37:14,920 --> 00:37:17,320 around the time of Kansas Styx, that time 1067 00:37:17,320 --> 00:37:20,840 frame, eighties, '70 late seventies, early eighties, there 1068 00:37:20,840 --> 00:37:22,139 was a lot of innovation 1069 00:37:22,440 --> 00:37:23,099 in music 1070 00:37:23,565 --> 00:37:24,784 around electronics, 1071 00:37:25,164 --> 00:37:26,764 and people could do a lot of really 1072 00:37:26,764 --> 00:37:27,505 cool stuff. 1073 00:37:27,804 --> 00:37:28,625 But the people 1074 00:37:29,085 --> 00:37:30,464 doing the cool stuff 1075 00:37:31,005 --> 00:37:33,025 grew up busting their fingers 1076 00:37:34,364 --> 00:37:37,425 on steel, you know, on steel guitar strings, 1077 00:37:37,880 --> 00:37:40,440 and they still had the feeling of how 1078 00:37:40,440 --> 00:37:41,500 to create music. 1079 00:37:42,039 --> 00:37:44,519 And so they used the tools to create 1080 00:37:44,519 --> 00:37:45,019 music. 1081 00:37:45,719 --> 00:37:47,739 When that generation went away, 1082 00:37:48,840 --> 00:37:51,260 modern music to me is all just bland. 1083 00:37:52,215 --> 00:37:54,375 Every song on the radio sounds the same 1084 00:37:54,375 --> 00:37:57,195 to me. It's all developed with the same 1085 00:37:58,135 --> 00:37:58,635 stupid 1086 00:37:59,175 --> 00:38:00,235 whatever whatever. 1087 00:38:00,775 --> 00:38:01,515 And so 1088 00:38:01,894 --> 00:38:03,815 there comes a point where I wonder about 1089 00:38:03,815 --> 00:38:04,954 that network engineering. 1090 00:38:05,550 --> 00:38:07,070 Like, at some point, are we gonna lose 1091 00:38:07,070 --> 00:38:08,930 the people who develop the protocols, 1092 00:38:09,949 --> 00:38:10,769 who fought 1093 00:38:11,230 --> 00:38:12,210 with the constraints, 1094 00:38:13,390 --> 00:38:15,550 and then develop tools to get around the 1095 00:38:15,550 --> 00:38:16,050 constraints. 1096 00:38:17,230 --> 00:38:19,550 And now the tools run the network and 1097 00:38:19,550 --> 00:38:20,449 not the person. 1098 00:38:21,215 --> 00:38:23,295 And and that that's a weird I don't 1099 00:38:23,295 --> 00:38:25,695 know. One step from being the angry guy 1100 00:38:25,695 --> 00:38:27,554 shaking a cloud shaking a cloud. 1101 00:38:29,614 --> 00:38:31,155 Get off my lawn. 1102 00:38:31,775 --> 00:38:33,695 Well and I think, like, I'm I'm kind 1103 00:38:33,695 --> 00:38:35,875 of a pragmatist when it comes to those 1104 00:38:36,969 --> 00:38:39,289 challenges. I mean, so yes, that may be 1105 00:38:39,289 --> 00:38:40,829 true. That may be happening. 1106 00:38:41,450 --> 00:38:42,989 Yes. I mean, you look at architecture. 1107 00:38:43,530 --> 00:38:45,530 Right? I mean, there was there was so 1108 00:38:45,530 --> 00:38:47,309 much richness in architecture. 1109 00:38:47,930 --> 00:38:48,829 Oh, yeah. Probably 1110 00:38:49,210 --> 00:38:50,570 the same way. 500 1111 00:38:50,570 --> 00:38:53,265 ago. Right? Even a hundred years ago. 1112 00:38:54,125 --> 00:38:56,545 Yeah. Like, it's not the same now. 1113 00:38:57,244 --> 00:38:57,744 Yep. 1114 00:38:58,285 --> 00:38:58,785 Like, 1115 00:39:01,085 --> 00:39:02,844 some of these things I think are just 1116 00:39:02,844 --> 00:39:04,844 the march of time and innovation and, like, 1117 00:39:04,844 --> 00:39:06,449 there's just nothing we can do about it. 1118 00:39:06,690 --> 00:39:07,190 Right? 1119 00:39:08,210 --> 00:39:09,809 I don't know. I don't I don't like, 1120 00:39:09,809 --> 00:39:12,130 is that a solvable problem? I don't know. 1121 00:39:12,130 --> 00:39:13,889 Because I I feel like we've got plenty 1122 00:39:13,889 --> 00:39:15,349 of examples that say 1123 00:39:16,130 --> 00:39:18,849 there the the creativity will move to a 1124 00:39:18,849 --> 00:39:20,869 different place of of humanity. 1125 00:39:21,170 --> 00:39:23,325 And, you know, I don't know. 1126 00:39:23,785 --> 00:39:25,505 Maybe it's just the market. I hope you're 1127 00:39:25,505 --> 00:39:27,704 I hope you're right, Yvonne. I'm kinda sanguine 1128 00:39:27,704 --> 00:39:29,864 about it because I don't that's like out 1129 00:39:29,864 --> 00:39:30,985 of my control. So 1130 00:39:31,785 --> 00:39:34,344 I think the instrumentation in networks today is 1131 00:39:34,344 --> 00:39:36,739 like so much better than it was before. 1132 00:39:37,760 --> 00:39:39,440 I think there's advanced I mean, the the 1133 00:39:39,440 --> 00:39:41,699 open source tooling that's available is 1134 00:39:42,239 --> 00:39:44,420 is way different. The the 1135 00:39:44,880 --> 00:39:46,640 the understanding of how to build an open 1136 00:39:46,640 --> 00:39:48,980 source practice inside of companies is 1137 00:39:49,704 --> 00:39:52,025 dramatically different than, you know, than even ten 1138 00:39:52,025 --> 00:39:53,565 years ago, fifteen years ago. 1139 00:39:54,264 --> 00:39:55,784 So I think a lot of stuff kinda 1140 00:39:55,784 --> 00:39:57,484 comes along, you know, the 1141 00:39:58,664 --> 00:39:59,164 the 1142 00:40:00,105 --> 00:40:01,784 I I I so I kind of agree 1143 00:40:01,784 --> 00:40:03,625 and kinda and kinda I'm not sure, but 1144 00:40:03,625 --> 00:40:04,605 maybe I'm more hopeful. 1145 00:40:05,440 --> 00:40:07,280 And the thing that gives me fear is 1146 00:40:07,280 --> 00:40:09,440 that people if if you relegate the network 1147 00:40:09,440 --> 00:40:11,280 to electricity where you don't, you know, even 1148 00:40:11,519 --> 00:40:13,199 like, no one turns lights on anymore. You 1149 00:40:13,199 --> 00:40:15,059 just walk into a room and lights happen. 1150 00:40:16,079 --> 00:40:17,679 If you relegate it to that kind of 1151 00:40:17,679 --> 00:40:18,980 utility, then I think 1152 00:40:19,565 --> 00:40:21,244 the number of people who pay attention to 1153 00:40:21,244 --> 00:40:23,325 it is is fewer, and and that's a 1154 00:40:23,325 --> 00:40:23,825 harder 1155 00:40:24,364 --> 00:40:25,724 like, that's the part I worry about is 1156 00:40:25,724 --> 00:40:26,625 you could drain 1157 00:40:27,005 --> 00:40:27,744 the workforce. 1158 00:40:29,765 --> 00:40:31,164 I but I do think that the the 1159 00:40:31,164 --> 00:40:33,869 workforce that's there with the the tooling that 1160 00:40:33,869 --> 00:40:34,609 they have, 1161 00:40:35,469 --> 00:40:37,089 it does give them more 1162 00:40:37,869 --> 00:40:40,349 capability than they otherwise. They're relying more on 1163 00:40:40,349 --> 00:40:42,989 the car, maybe on the tooling. Maybe it's 1164 00:40:42,989 --> 00:40:44,349 more like a, you know, stick shift in 1165 00:40:44,349 --> 00:40:45,010 an automatic, 1166 00:40:45,804 --> 00:40:47,804 like people can still drive cars. They just, 1167 00:40:47,804 --> 00:40:50,204 maybe they couldn't drive a stick, but the, 1168 00:40:50,204 --> 00:40:51,505 but the, they're 1169 00:40:52,045 --> 00:40:54,285 still able to make a car work. Maybe 1170 00:40:54,285 --> 00:40:55,324 that's where we end up. 1171 00:40:55,965 --> 00:40:57,885 It's an interesting question, how many people you 1172 00:40:57,885 --> 00:41:00,304 need to actually, know the underlying bits. 1173 00:41:00,619 --> 00:41:02,380 And then as you plan your workforce, then 1174 00:41:02,380 --> 00:41:03,920 the question would be, what's the ratio? 1175 00:41:04,380 --> 00:41:06,059 Like how many experts should you have on 1176 00:41:06,059 --> 00:41:08,539 staff? And how many, you know, generalists, you 1177 00:41:08,539 --> 00:41:10,480 know, this is a specialist in general generalist 1178 00:41:10,619 --> 00:41:12,780 conversation again. You know, what's the right ratio? 1179 00:41:12,780 --> 00:41:14,000 And so if you're planning, 1180 00:41:14,445 --> 00:41:15,965 maybe people should look at that and say, 1181 00:41:15,965 --> 00:41:17,184 let me get my hands on 1182 00:41:17,485 --> 00:41:19,505 the the three specialists I need now. 1183 00:41:19,805 --> 00:41:21,485 Let me make sure they're comped and and 1184 00:41:21,485 --> 00:41:23,325 good to go. And then let me set 1185 00:41:23,325 --> 00:41:24,844 up let me hire the rest, and then 1186 00:41:24,844 --> 00:41:27,405 let me put together a mentoring plan so 1187 00:41:27,405 --> 00:41:29,670 my three specialists are always mentoring the next 1188 00:41:29,670 --> 00:41:30,650 three that follow. 1189 00:41:31,349 --> 00:41:33,590 Yeah. Yeah. And this is all go this 1190 00:41:33,590 --> 00:41:34,969 still goes back to 1191 00:41:35,510 --> 00:41:38,469 be living in a constrained environment. Right? This 1192 00:41:38,469 --> 00:41:40,789 all still goes back to how do you 1193 00:41:40,789 --> 00:41:41,289 manage 1194 00:41:41,750 --> 00:41:42,710 things when you're in a 1195 00:41:47,804 --> 00:41:50,204 constraints in hardware, we are in in just 1196 00:41:50,284 --> 00:41:51,984 but we also face constraints 1197 00:41:52,284 --> 00:41:53,344 in in expertise, 1198 00:41:53,804 --> 00:41:54,304 in 1199 00:41:54,924 --> 00:41:55,904 in just knowledge. 1200 00:41:56,525 --> 00:41:58,144 And this, like, this intersection 1201 00:41:58,980 --> 00:42:00,760 of constraints that are causing 1202 00:42:01,619 --> 00:42:03,320 a lot of heartache right now. 1203 00:42:04,179 --> 00:42:06,980 The workforce constraints are are just as pronounced 1204 00:42:06,980 --> 00:42:09,300 as the the hardware constraints. Everyone right now 1205 00:42:09,300 --> 00:42:10,920 trying to build out GPU clusters. 1206 00:42:11,380 --> 00:42:12,739 The number of people, you know, it's like, 1207 00:42:12,739 --> 00:42:14,019 oh, I want someone who's got, you know, 1208 00:42:14,019 --> 00:42:15,404 five years experience with GPU clusters. 1209 00:42:16,744 --> 00:42:18,184 Like, go ahead. There are 10 of them 1210 00:42:18,184 --> 00:42:20,105 in the world. Right? Like, you know, it's 1211 00:42:20,105 --> 00:42:21,644 something like that. Yeah. 1212 00:42:23,144 --> 00:42:24,525 Yeah. It's it's it's insane. 1213 00:42:24,904 --> 00:42:25,644 And again, 1214 00:42:26,344 --> 00:42:28,940 this is where the generals comes back in 1215 00:42:28,940 --> 00:42:29,599 and says, 1216 00:42:29,900 --> 00:42:31,900 but I know about networks. I know how 1217 00:42:31,900 --> 00:42:33,279 to build data center fabrics. 1218 00:42:33,739 --> 00:42:35,579 I can pick up how to build a 1219 00:42:35,579 --> 00:42:36,719 GPU cluster 1220 00:42:37,739 --> 00:42:39,440 by just adding some bits. 1221 00:42:39,819 --> 00:42:41,599 I I can figure that out. 1222 00:42:41,925 --> 00:42:44,005 You don't need somebody who's done it for 1223 00:42:44,005 --> 00:42:45,864 five years or ten years exclusively 1224 00:42:46,565 --> 00:42:48,344 for probably 80% 1225 00:42:48,805 --> 00:42:50,644 of the GP clusters that need to be 1226 00:42:50,644 --> 00:42:51,144 built. 1227 00:42:52,805 --> 00:42:54,565 Maybe. I don't know. What do you think 1228 00:42:54,565 --> 00:42:56,105 of that, Mike? Is that a crazy? 1229 00:42:57,880 --> 00:42:59,719 I think I could probably get to the 1230 00:42:59,719 --> 00:43:01,440 same conclusion, but a different way. I just 1231 00:43:01,559 --> 00:43:02,699 I don't think that 1232 00:43:03,880 --> 00:43:05,559 and this is, I don't I don't mean 1233 00:43:05,559 --> 00:43:07,579 to wax philosophical on this, but, 1234 00:43:08,280 --> 00:43:09,500 like, I I think companies 1235 00:43:10,005 --> 00:43:11,525 are with with the whole AI thing in 1236 00:43:11,525 --> 00:43:13,385 particular, I think there's a lot of companies, 1237 00:43:13,444 --> 00:43:15,045 a lot of or maybe a lot of, 1238 00:43:15,045 --> 00:43:16,804 you know, execs that are looking to say, 1239 00:43:16,804 --> 00:43:17,545 how do I 1240 00:43:18,085 --> 00:43:20,324 effectively reduce a bunch of cost, and how 1241 00:43:20,324 --> 00:43:22,085 do I, you know, usher people out the 1242 00:43:22,085 --> 00:43:23,880 door? I think there needs to be, like, 1243 00:43:23,880 --> 00:43:25,639 a some loyalty there, and people have to 1244 00:43:25,639 --> 00:43:28,199 say, I'm committed to bringing people along. And 1245 00:43:28,199 --> 00:43:29,719 I think if you take someone who's motivated 1246 00:43:29,719 --> 00:43:31,400 and you put them in an environment where 1247 00:43:31,400 --> 00:43:32,940 the conditions are are healthy, 1248 00:43:33,719 --> 00:43:35,500 they can make the transitions. 1249 00:43:35,800 --> 00:43:37,880 I think companies should invest in that. I 1250 00:43:37,880 --> 00:43:40,355 think, one, it's a practical reality. There's not 1251 00:43:40,355 --> 00:43:42,675 enough people, not enough resources, generally speaking, that 1252 00:43:42,675 --> 00:43:43,815 have the kind of experience. 1253 00:43:44,434 --> 00:43:45,974 Two, I think that technology 1254 00:43:46,514 --> 00:43:49,414 in general is interesting. Technology contextualized 1255 00:43:50,275 --> 00:43:52,034 is difference making, and I think that's the 1256 00:43:52,034 --> 00:43:52,534 contextualization 1257 00:43:52,914 --> 00:43:54,890 that matters. And so there's it's in your 1258 00:43:54,890 --> 00:43:56,670 best interest anyway to do that. 1259 00:43:57,210 --> 00:43:58,410 And then I think if you do that, 1260 00:43:58,410 --> 00:44:00,250 then you can you can you can still 1261 00:44:00,250 --> 00:44:02,090 do staff augmentation, but I don't think you 1262 00:44:02,090 --> 00:44:04,090 need quite as many people. And then just 1263 00:44:04,090 --> 00:44:06,090 practically speaking, if you look at the economics, 1264 00:44:06,090 --> 00:44:08,170 if there's very few people who understand a 1265 00:44:08,170 --> 00:44:08,670 particular 1266 00:44:09,974 --> 00:44:12,074 skill or who have a particular skill, 1267 00:44:13,335 --> 00:44:15,414 then they're gonna be in crazy demand. And 1268 00:44:15,414 --> 00:44:17,414 so if you don't have the budget to 1269 00:44:17,414 --> 00:44:20,154 compete with a a Microsoft or an Oracle 1270 00:44:20,295 --> 00:44:21,355 or a or a Meta, 1271 00:44:22,464 --> 00:44:22,964 then 1272 00:44:23,579 --> 00:44:25,420 maybe what you should be doing is is 1273 00:44:25,420 --> 00:44:27,739 building out your own team and and, you 1274 00:44:27,739 --> 00:44:29,820 know, bringing people along. That's what I would 1275 00:44:29,820 --> 00:44:31,579 do. And so as you as you think 1276 00:44:31,579 --> 00:44:33,760 through, like, the the strategies, then it's really, 1277 00:44:34,059 --> 00:44:35,660 you know, again, going back to some of 1278 00:44:35,660 --> 00:44:37,420 the abstraction comments, it's like, you know, what's 1279 00:44:37,420 --> 00:44:39,255 the level of expertise I need? 1280 00:44:39,654 --> 00:44:42,154 You know, how many specialists do I need? 1281 00:44:42,454 --> 00:44:44,534 You know, what what am I optimizing for? 1282 00:44:44,534 --> 00:44:45,514 What are the architectural 1283 00:44:45,894 --> 00:44:48,375 concessions I'm willing to make? And then what 1284 00:44:48,375 --> 00:44:50,614 are the and then how much optionality do 1285 00:44:50,614 --> 00:44:53,309 I wanna maintain? How many sources of whatever 1286 00:44:53,309 --> 00:44:54,750 do I wanna have? And then use that 1287 00:44:54,750 --> 00:44:55,410 to dictate, 1288 00:44:55,710 --> 00:44:57,329 you know, where you where you go. 1289 00:44:57,869 --> 00:45:00,269 But that's very different than, you know, new 1290 00:45:00,269 --> 00:45:02,130 new project comes in, you know, 1291 00:45:02,510 --> 00:45:04,454 pull out the old RFP, you know, add 1292 00:45:04,454 --> 00:45:06,375 three lines to the Excel file and then, 1293 00:45:06,375 --> 00:45:07,974 you know, send it back out and and 1294 00:45:07,974 --> 00:45:09,275 and work on the next RFP. 1295 00:45:09,655 --> 00:45:11,255 And how do I keep people once I've 1296 00:45:11,255 --> 00:45:11,994 trained them? 1297 00:45:12,775 --> 00:45:14,614 And that's and that's and I'm just saying 1298 00:45:14,614 --> 00:45:16,775 that's something you gotta think about. Right? That's 1299 00:45:16,775 --> 00:45:17,835 always an issue. 1300 00:45:18,570 --> 00:45:20,010 Once I put people in the past That's 1301 00:45:20,010 --> 00:45:21,610 a whole podcast, man. Like, I know. I 1302 00:45:21,610 --> 00:45:23,210 know. But it is but it's just something 1303 00:45:23,210 --> 00:45:23,869 to consider. 1304 00:45:24,250 --> 00:45:26,170 I think you gotta pay people what they're 1305 00:45:26,170 --> 00:45:28,670 what they're worth before they ask for it. 1306 00:45:30,250 --> 00:45:32,489 You create an environment where they wanna be 1307 00:45:32,489 --> 00:45:32,989 and 1308 00:45:33,545 --> 00:45:36,025 keep other people from stealing them, which, like, 1309 00:45:36,025 --> 00:45:38,025 people are only gonna steal somebody away from 1310 00:45:38,025 --> 00:45:40,184 where they wanna be if you're you're underpaying 1311 00:45:40,184 --> 00:45:41,864 them. Like, if you pay them what they're 1312 00:45:41,864 --> 00:45:43,864 worth and you put them in an environment 1313 00:45:43,864 --> 00:45:45,325 where they're reasonably happy, 1314 00:45:45,719 --> 00:45:46,460 like, generally, 1315 00:45:47,000 --> 00:45:49,099 they're not gonna move. Like, 1316 00:45:49,880 --> 00:45:52,039 I know that sounds super basic. But Oh 1317 00:45:52,039 --> 00:45:54,280 my god. That that if you're willing to 1318 00:45:54,280 --> 00:45:55,019 pay somebody 1319 00:45:55,880 --> 00:45:58,840 after they've raised their, their their their counter 1320 00:45:58,840 --> 00:46:00,994 or their offer from someone else, If you're 1321 00:46:00,994 --> 00:46:02,195 willing to pay them then, then you should 1322 00:46:02,195 --> 00:46:03,875 be willing to pay pay them without the 1323 00:46:03,875 --> 00:46:04,375 offer. 1324 00:46:04,755 --> 00:46:06,195 Like, oh my god. Do right by your 1325 00:46:06,195 --> 00:46:08,355 people. I'll I won't go deeper on that, 1326 00:46:08,355 --> 00:46:10,114 but but Yeah. Well, maybe it's a good 1327 00:46:10,114 --> 00:46:11,555 show to have. Maybe we should talk about 1328 00:46:11,555 --> 00:46:13,875 it sometime. Even though I'll just say that 1329 00:46:13,875 --> 00:46:16,489 there is a con currently a constraint around 1330 00:46:16,489 --> 00:46:18,489 the kind of thinking that Mike Bouchon just 1331 00:46:18,489 --> 00:46:19,789 expressed on this podcast. 1332 00:46:20,329 --> 00:46:22,969 That's another constraint that we have. Yeah. Oh, 1333 00:46:22,969 --> 00:46:24,030 yes. Of course. 1334 00:46:24,650 --> 00:46:26,650 Alright. Cool. Well, I think we've hit everything 1335 00:46:26,650 --> 00:46:29,049 unless you have other things you wanna say 1336 00:46:29,049 --> 00:46:31,835 there, Mike. Are you kinda I'm good. How 1337 00:46:31,835 --> 00:46:32,894 about you, Tom? 1338 00:46:33,194 --> 00:46:35,295 No. I'm good. And you, Yvonne? 1339 00:46:35,755 --> 00:46:38,315 All good. Let's call it the weekend. Let's 1340 00:46:38,315 --> 00:46:40,574 call it the weekend. It is Friday afternoon. 1341 00:46:40,875 --> 00:46:41,375 Alright. 1342 00:46:42,154 --> 00:46:44,154 So, Tom, where can people get in touch 1343 00:46:44,154 --> 00:46:45,674 with you? If they want to, say your 1344 00:46:45,674 --> 00:46:46,174 word. 1345 00:46:48,239 --> 00:46:48,739 LinkedIn. 1346 00:46:49,840 --> 00:46:51,300 Did you time that one? 1347 00:46:51,599 --> 00:46:53,539 I did not time it. No? 1348 00:46:53,840 --> 00:46:56,159 I'll I'll bring the stopwatch out. We'll time 1349 00:46:56,159 --> 00:46:57,760 them in the future and figure out what's 1350 00:46:57,760 --> 00:47:00,159 going on with the with the variable length 1351 00:47:00,159 --> 00:47:01,619 pause. And Yvonne? 1352 00:47:02,664 --> 00:47:03,144 Yeah. 1353 00:47:03,464 --> 00:47:06,045 LinkedIn and Yvonne Sharp or on, 1354 00:47:06,585 --> 00:47:09,385 Twitter, x at Sharp Network. I'm I'm lurking 1355 00:47:09,385 --> 00:47:11,625 more than I'm talking these days, but I 1356 00:47:11,625 --> 00:47:12,525 am still around. 1357 00:47:13,304 --> 00:47:14,039 And Mike. 1358 00:47:14,440 --> 00:47:17,000 I'm also on LinkedIn these days. I gave 1359 00:47:17,000 --> 00:47:17,900 up the socials 1360 00:47:18,519 --> 00:47:20,300 for the tire fire that they are. 1361 00:47:22,039 --> 00:47:24,300 Okay. Man. Yep. Cool. 1362 00:47:25,000 --> 00:47:26,920 Alright. And I'm Russ White. You can find 1363 00:47:26,920 --> 00:47:27,500 me here 1364 00:47:28,494 --> 00:47:29,934 the hedge. You can find me at rule 1365 00:47:29,934 --> 00:47:31,775 eleven dot tech. You can find me on 1366 00:47:31,775 --> 00:47:32,275 LinkedIn. 1367 00:47:32,974 --> 00:47:34,894 I haven't logged into x in a while, 1368 00:47:34,894 --> 00:47:36,815 but, well, you know, if you PM me, 1369 00:47:36,815 --> 00:47:39,135 I'll probably eventually see it over there. Maybe. 1370 00:47:39,135 --> 00:47:39,875 Who knows? 1371 00:47:40,349 --> 00:47:40,849 And 1372 00:47:41,150 --> 00:47:42,989 we know that your attention is important. We 1373 00:47:42,989 --> 00:47:44,829 live in an attention driven economy, so we 1374 00:47:44,829 --> 00:47:47,550 thank you for listening all the way to 1375 00:47:47,550 --> 00:47:49,309 the bitter end of this episode of The 1376 00:47:49,309 --> 00:47:51,250 Hedge, and we will catch you next time.