1 00:00:01,760 --> 00:00:03,839 Join us as we gather around the hedge, 2 00:00:03,839 --> 00:00:05,219 where we dig into technology, 3 00:00:05,519 --> 00:00:08,240 business, and culture with the finest minds in 4 00:00:08,240 --> 00:00:09,220 computer networking. 5 00:00:20,904 --> 00:00:23,324 Well, hello, Anil. Where are you physically? 6 00:00:24,664 --> 00:00:26,664 Hi, Russ. Thanks for having me. I am 7 00:00:26,664 --> 00:00:28,125 in San Francisco, California. 8 00:00:28,600 --> 00:00:29,500 San Francisco. 9 00:00:30,039 --> 00:00:32,300 Oh, cool. I used to work for LinkedIn. 10 00:00:32,600 --> 00:00:34,679 So I know San Francisco well because their 11 00:00:34,679 --> 00:00:36,859 headquarters is down on 222 Second. 12 00:00:37,240 --> 00:00:39,020 And I used to walk from, 13 00:00:40,120 --> 00:00:40,780 the Hilton 14 00:00:41,320 --> 00:00:41,820 on 15 00:00:42,554 --> 00:00:44,634 right up by Eighth Street. I don't remember 16 00:00:44,634 --> 00:00:47,515 the cross street there, but and then I 17 00:00:47,515 --> 00:00:50,155 walked down Eighth or Seventh or Fourth. Maybe 18 00:00:50,155 --> 00:00:51,695 it's Fourth or something I'd hit. 19 00:00:52,475 --> 00:00:53,934 And I'd run down Market 20 00:00:54,234 --> 00:00:55,394 to get to work in the morning when 21 00:00:55,394 --> 00:00:56,859 I was staying in the Hilton. And I 22 00:00:56,859 --> 00:00:58,939 would stop at the Ghirardelli for for hot 23 00:00:58,939 --> 00:01:00,320 chocolate because that's like 24 00:01:01,100 --> 00:01:02,079 good hot chocolate. 25 00:01:05,019 --> 00:01:07,180 That's great. And then before I left, I 26 00:01:07,180 --> 00:01:09,340 always picked up something from the moleskin store 27 00:01:09,340 --> 00:01:11,680 because my wife loves notebooks, 28 00:01:12,474 --> 00:01:14,715 loves notebooks and pens. So I was always 29 00:01:14,715 --> 00:01:17,215 going to the moleskin store going, okay. What 30 00:01:17,354 --> 00:01:19,274 pen have a dye bought her from the 31 00:01:19,274 --> 00:01:20,414 moleskin store? 32 00:01:21,194 --> 00:01:22,894 Because it's right there on Market. 33 00:01:23,194 --> 00:01:23,694 Yeah. 34 00:01:24,474 --> 00:01:24,974 So, 35 00:01:26,170 --> 00:01:27,930 I guess, maybe start here, and you'll tell 36 00:01:27,930 --> 00:01:29,930 me a little bit about yourself. Right now, 37 00:01:29,930 --> 00:01:31,370 we don't have any we don't have Tom. 38 00:01:31,370 --> 00:01:33,049 We don't have Yvonne. So it's just Anil 39 00:01:33,049 --> 00:01:35,209 and I for our regular listeners. You know, 40 00:01:35,209 --> 00:01:37,609 you can be disappointed and hang up now 41 00:01:37,609 --> 00:01:39,370 if you want to and stop listening because 42 00:01:39,370 --> 00:01:41,644 we don't have Any of the interesting people 43 00:01:41,644 --> 00:01:44,045 on. But yeah. So talk to me a 44 00:01:44,045 --> 00:01:45,885 little bit about yourself, Anil. Like, what's your 45 00:01:45,885 --> 00:01:48,385 background and kind of what you're working on? 46 00:01:49,084 --> 00:01:49,584 Sure. 47 00:01:50,045 --> 00:01:52,224 My background is I studied network engineering, 48 00:01:53,165 --> 00:01:55,025 built networking systems for about 49 00:01:55,409 --> 00:01:58,210 a decade at a previous company, and then 50 00:01:58,210 --> 00:02:00,230 I helped start Meter about 51 00:02:00,850 --> 00:02:02,469 nine, ten years ago now. 52 00:02:03,969 --> 00:02:06,870 Meter is a networking company. 53 00:02:07,730 --> 00:02:08,710 We build, 54 00:02:09,545 --> 00:02:11,004 deploy, and maintain 55 00:02:12,344 --> 00:02:12,844 networking 56 00:02:13,385 --> 00:02:14,525 hardware and software. 57 00:02:15,064 --> 00:02:15,564 Okay. 58 00:02:15,944 --> 00:02:16,764 We design 59 00:02:17,705 --> 00:02:20,105 everything from the hardware on up. So this 60 00:02:20,105 --> 00:02:21,240 is everything for 61 00:02:22,120 --> 00:02:24,860 routing, switching, wireless, security. 62 00:02:25,240 --> 00:02:26,760 I know, Russ, you can see our rack 63 00:02:26,760 --> 00:02:29,800 behind us. Yeah. Yeah. But it's it's our 64 00:02:29,800 --> 00:02:30,919 rack where it's our, 65 00:02:31,400 --> 00:02:31,900 PDUs, 66 00:02:33,735 --> 00:02:34,235 firewalls, 67 00:02:35,094 --> 00:02:35,594 switches, 68 00:02:36,375 --> 00:02:38,555 indoor Wi Fi, outdoor Wi Fi, cellular, 69 00:02:39,335 --> 00:02:41,495 where we design and build the hardware, like 70 00:02:41,495 --> 00:02:44,314 the firmware, operating systems, distributed systems, APIs, 71 00:02:44,775 --> 00:02:47,094 applications. So basically a full en route switch 72 00:02:47,094 --> 00:02:49,199 company that does hardware all the way up. 73 00:02:49,520 --> 00:02:52,400 And I Hardware. Yes. And and we work 74 00:02:52,400 --> 00:02:54,479 closely with our customers and partners to help 75 00:02:54,479 --> 00:02:56,719 deploy and maintain it as well. Okay. 76 00:02:57,519 --> 00:02:59,599 And what we do differently sort of three 77 00:02:59,599 --> 00:03:01,919 things in all these years of networking is, 78 00:03:01,919 --> 00:03:04,125 you know, Russ, majority of the networking stacks 79 00:03:04,205 --> 00:03:05,585 have been built through acquisition. 80 00:03:05,965 --> 00:03:06,465 Mhmm. 81 00:03:07,645 --> 00:03:09,724 We sort of decided that there should be 82 00:03:09,724 --> 00:03:12,125 a company. Ironically, we're the first people to 83 00:03:12,125 --> 00:03:14,284 do the entire rack together with a single 84 00:03:14,284 --> 00:03:15,264 operating system, 85 00:03:15,724 --> 00:03:18,444 single firmware, a single API for the entire 86 00:03:18,444 --> 00:03:18,900 rack, 87 00:03:19,860 --> 00:03:23,459 and fully integrated products across ISP management, routing, 88 00:03:23,459 --> 00:03:24,519 switching, wireless, 89 00:03:25,379 --> 00:03:25,879 SDN, 90 00:03:26,180 --> 00:03:27,159 DNS security, 91 00:03:27,539 --> 00:03:28,039 VPN, 92 00:03:28,419 --> 00:03:30,919 IDS, IPS, all in a single platform. 93 00:03:32,165 --> 00:03:34,724 Then we're vertically integrated in that. We also 94 00:03:34,724 --> 00:03:37,284 help deploy these networks everywhere in the world 95 00:03:37,284 --> 00:03:39,764 with our customers and partners. We work with 96 00:03:39,764 --> 00:03:41,145 the myriad of channel partners. 97 00:03:41,844 --> 00:03:43,605 And then our pricing model is a bit 98 00:03:43,605 --> 00:03:45,844 different. We take on all the risk for 99 00:03:45,844 --> 00:03:46,504 the hardware. 100 00:03:46,965 --> 00:03:48,639 So there's there's no capital costs. 101 00:03:49,659 --> 00:03:52,379 Okay. And when there's existing hardware in place, 102 00:03:52,379 --> 00:03:54,300 we'll buy that out. And then when we 103 00:03:54,300 --> 00:03:55,280 build new hardware, 104 00:03:55,580 --> 00:03:57,900 we'll upgrade our customers for free as well. 105 00:03:57,900 --> 00:03:58,400 So 106 00:03:59,019 --> 00:04:00,860 so this is a networking company from the 107 00:04:00,860 --> 00:04:03,564 ground up. Cool. So this is all very 108 00:04:03,564 --> 00:04:06,525 operationally oriented. That's what I find interesting about 109 00:04:06,525 --> 00:04:08,944 it, is that you're not focused 110 00:04:09,805 --> 00:04:11,825 so much on the route, 111 00:04:12,444 --> 00:04:14,784 routing features, speeds and feeds, 112 00:04:15,379 --> 00:04:17,139 stuff like that. I'm sure you have plenty 113 00:04:17,139 --> 00:04:19,300 of features, but that's not really your primary 114 00:04:19,300 --> 00:04:21,560 focus. Your primary focus is on operations, 115 00:04:22,019 --> 00:04:24,180 which takes us back to a conversation we 116 00:04:24,180 --> 00:04:25,240 had with Mike Bouchong, 117 00:04:26,420 --> 00:04:27,240 on the hedge 118 00:04:27,540 --> 00:04:28,759 not too long ago, 119 00:04:29,185 --> 00:04:30,865 maybe a couple of months. I always forget 120 00:04:30,865 --> 00:04:32,225 when these things post, so I don't I 121 00:04:32,225 --> 00:04:34,245 don't remember how long it's been since I've 122 00:04:34,305 --> 00:04:35,845 talked to somebody about something. 123 00:04:36,544 --> 00:04:37,044 That 124 00:04:37,425 --> 00:04:39,745 operations is kind of the future or is 125 00:04:39,745 --> 00:04:41,444 the future of network engineering. 126 00:04:42,060 --> 00:04:43,660 We were round and about about that a 127 00:04:43,660 --> 00:04:45,279 couple of times with Mike, and, 128 00:04:45,660 --> 00:04:47,500 I pushed back on him on some things. 129 00:04:47,500 --> 00:04:49,339 And he pushed back on me, which is 130 00:04:49,339 --> 00:04:50,319 all good because 131 00:04:50,860 --> 00:04:52,300 Mike and I are good friends, so that's 132 00:04:52,300 --> 00:04:53,100 good. So, 133 00:04:54,754 --> 00:04:58,034 I guess if you're so operations focused, it's 134 00:04:58,034 --> 00:05:00,194 interesting the tack you're taking or the path 135 00:05:00,194 --> 00:05:03,095 you're taking is unified operating system. 136 00:05:03,634 --> 00:05:05,634 Now are you writing your own stacks, or 137 00:05:05,634 --> 00:05:07,875 are you using open source like FR routing 138 00:05:07,875 --> 00:05:09,509 or BER or something like that? 139 00:05:09,990 --> 00:05:12,150 Everything is us from the ground up. And 140 00:05:12,150 --> 00:05:14,790 okay, maybe to correct you just a slight 141 00:05:14,790 --> 00:05:17,110 bit. We do care about the speeds, we 142 00:05:17,110 --> 00:05:18,490 do care about the speeds, 143 00:05:18,870 --> 00:05:20,710 we do care about every inch of the 144 00:05:20,710 --> 00:05:21,770 hardware, we 145 00:05:22,545 --> 00:05:25,345 are care designing the hardware. Our incentives are 146 00:05:25,345 --> 00:05:27,264 very different because if you look at all 147 00:05:27,264 --> 00:05:30,305 these legacy companies, they are building hardware just 148 00:05:30,305 --> 00:05:32,064 so they can build it for as cheaply 149 00:05:32,064 --> 00:05:32,725 as possible 150 00:05:33,105 --> 00:05:35,120 and sell it for as expensive as possible. 151 00:05:35,180 --> 00:05:38,699 Our incentives are the hardware has to be 152 00:05:38,699 --> 00:05:41,180 great operationally like you're saying, because we're deploying 153 00:05:41,180 --> 00:05:42,860 it and it has to work really well 154 00:05:42,860 --> 00:05:44,620 because we only get paid if the service 155 00:05:44,620 --> 00:05:45,279 is great. 156 00:05:45,979 --> 00:05:48,060 So everything is in that light where we're 157 00:05:48,060 --> 00:05:49,979 incentivized to do the same thing with our 158 00:05:49,979 --> 00:05:50,324 customers. 159 00:05:51,045 --> 00:05:53,125 Customers. So the reason we also care about 160 00:05:53,125 --> 00:05:55,764 writing our own firmware, routing stack, all of 161 00:05:55,764 --> 00:05:56,264 it, 162 00:05:56,725 --> 00:05:59,865 is to make sure the operations and maintenance 163 00:06:00,324 --> 00:06:01,225 are spectacular. 164 00:06:03,180 --> 00:06:05,899 So the network can run really smoothly and 165 00:06:05,899 --> 00:06:06,879 not be a challenge. 166 00:06:07,339 --> 00:06:08,939 But at the same time, we're not trying 167 00:06:08,939 --> 00:06:11,360 to take away any control from network engineers. 168 00:06:11,899 --> 00:06:14,779 The software we invest in is heavily down 169 00:06:14,779 --> 00:06:17,115 to being able to change, you know, like 170 00:06:17,115 --> 00:06:19,435 an STP priority on a port all the 171 00:06:19,435 --> 00:06:21,214 way to anything they'd like to change. 172 00:06:21,675 --> 00:06:23,855 We believe in great software still for, 173 00:06:24,235 --> 00:06:26,475 folks that are using our networks. So the 174 00:06:26,475 --> 00:06:29,115 incentive question is something that we face all 175 00:06:29,115 --> 00:06:31,055 the time in the network engineering world 176 00:06:31,355 --> 00:06:32,095 in that 177 00:06:32,620 --> 00:06:33,680 if you're incented 178 00:06:34,060 --> 00:06:36,459 to sell, and then basically, I don't want 179 00:06:36,459 --> 00:06:37,839 to say abandoned, but 180 00:06:38,300 --> 00:06:41,019 kind of, it's not your problem, you're not 181 00:06:41,019 --> 00:06:42,620 the one who gets the call at 02:00 182 00:06:42,620 --> 00:06:44,779 in the morning, when the hardware or the 183 00:06:44,779 --> 00:06:45,680 software fails, 184 00:06:46,185 --> 00:06:48,524 your incentives tend to be far different 185 00:06:49,145 --> 00:06:50,685 than if you are 186 00:06:51,064 --> 00:06:53,705 actually the company maintaining. If you deploy it 187 00:06:53,705 --> 00:06:55,625 and then you have to maintain it. And 188 00:06:55,625 --> 00:06:57,485 this is a point of friction in operations 189 00:06:57,545 --> 00:06:58,845 I see all the time. 190 00:06:59,545 --> 00:07:01,389 And as a matter of fact, I see 191 00:07:01,389 --> 00:07:02,669 it not just in the heart in the 192 00:07:02,669 --> 00:07:05,649 hardware software world within the services world. Right? 193 00:07:05,870 --> 00:07:07,970 Like, if you're a cloud provider, 194 00:07:08,509 --> 00:07:10,349 and you want your service to be up 195 00:07:10,349 --> 00:07:13,709 all the time, but this service availability number 196 00:07:13,709 --> 00:07:16,095 is kind of an abstract thing that's out 197 00:07:16,095 --> 00:07:17,074 there someplace. 198 00:07:17,615 --> 00:07:19,454 It's not really me. It's like this out 199 00:07:19,454 --> 00:07:22,495 there number. I'm I'm available 99% of the 200 00:07:22,495 --> 00:07:24,974 time, whatever the number is, five nines, four 201 00:07:24,974 --> 00:07:27,074 nines, whatever it is, doesn't matter. But 202 00:07:28,939 --> 00:07:31,520 when my cloud service fails, 203 00:07:32,139 --> 00:07:35,040 it's not me that's losing business. 204 00:07:36,139 --> 00:07:38,959 It's my customers who are losing business. 205 00:07:39,500 --> 00:07:41,500 It's my customers who get the call at 206 00:07:41,500 --> 00:07:42,879 02:00 in the morning, 207 00:07:43,485 --> 00:07:45,745 which leads me to build things. 208 00:07:47,245 --> 00:07:49,024 I don't say cut corners, perhaps, 209 00:07:49,404 --> 00:07:50,384 but to say 210 00:07:51,485 --> 00:07:53,725 a lot more bailing wire and bubble gun 211 00:07:53,725 --> 00:07:54,704 and duct tape. 212 00:07:55,509 --> 00:07:57,370 Then if you actually were the person 213 00:07:57,750 --> 00:07:59,589 doing the work of getting the call at 214 00:07:59,589 --> 00:08:01,830 02:00 in the morning, because you know, your 215 00:08:01,830 --> 00:08:03,209 customers down at two. 216 00:08:03,509 --> 00:08:05,670 Well, my business hours are nine to five, 217 00:08:05,670 --> 00:08:07,129 call me when it's 09:00. 218 00:08:07,509 --> 00:08:08,810 Yeah, I've had that right. 219 00:08:10,264 --> 00:08:12,024 It's interesting because when I was in Cisco 220 00:08:12,024 --> 00:08:13,485 tech and escalation team, 221 00:08:14,504 --> 00:08:16,985 many, many years ago, we never had that 222 00:08:16,985 --> 00:08:17,485 attitude. 223 00:08:18,504 --> 00:08:20,264 When I was in tech, that was something 224 00:08:20,264 --> 00:08:21,944 that was great about the Cisco tech when 225 00:08:21,944 --> 00:08:23,545 I was there. I don't know how it 226 00:08:23,545 --> 00:08:25,689 is now. But when I was there, if 227 00:08:25,689 --> 00:08:27,210 you took the case, you're on the problem 228 00:08:27,210 --> 00:08:29,710 until it closed, and it was your problem. 229 00:08:30,250 --> 00:08:32,090 And if the cast if your customer had 230 00:08:32,090 --> 00:08:35,610 customers down, that was your problem, not their 231 00:08:35,610 --> 00:08:36,110 problem. 232 00:08:36,649 --> 00:08:37,149 Right? 233 00:08:37,915 --> 00:08:40,075 You were on the hook to solve the 234 00:08:40,075 --> 00:08:40,575 problem. 235 00:08:41,115 --> 00:08:43,915 Make their network work again. That means shipping 236 00:08:43,915 --> 00:08:47,434 hardware have at if that meant shipping hardware 237 00:08:47,434 --> 00:08:49,295 and figuring the money out later. 238 00:08:50,649 --> 00:08:52,970 That's okay, too. Like, oh, they need to 239 00:08:52,970 --> 00:08:53,930 use 7200 240 00:08:53,930 --> 00:08:55,389 to get them back on the air. 241 00:08:55,850 --> 00:08:57,529 We didn't sit around and argue about who 242 00:08:57,529 --> 00:08:58,830 was paying for it. 243 00:08:59,290 --> 00:09:01,210 Right? You know, we picked up the phone, 244 00:09:01,210 --> 00:09:02,110 shipped the box, 245 00:09:02,495 --> 00:09:04,254 and then the account team could go figure 246 00:09:04,254 --> 00:09:05,695 out who was going to pay for it. 247 00:09:05,695 --> 00:09:08,495 That's that's not, that was not our problem 248 00:09:08,495 --> 00:09:10,654 intact. Our problem was not the money side. 249 00:09:10,654 --> 00:09:12,595 Our problem was to get the problem fixed. 250 00:09:13,615 --> 00:09:16,574 So, yeah, so I think the networking industry 251 00:09:16,574 --> 00:09:18,829 has changed a lot since then. 252 00:09:19,529 --> 00:09:22,490 And nowadays, it's much more, you bought the 253 00:09:22,490 --> 00:09:24,909 box, you figure it out, have fun. 254 00:09:25,289 --> 00:09:27,769 You know, that's right. So I think it's 255 00:09:27,769 --> 00:09:30,190 an interesting model that you're proposing, 256 00:09:31,049 --> 00:09:33,309 just at least from an incentives perspective. 257 00:09:34,995 --> 00:09:38,274 Now, from a network reliability perspective, something you 258 00:09:38,274 --> 00:09:39,975 said is very interesting to me too, 259 00:09:41,794 --> 00:09:43,794 that you're still trying to get deep into 260 00:09:43,794 --> 00:09:44,294 features. 261 00:09:45,075 --> 00:09:46,995 And maybe I'll push back on this a 262 00:09:46,995 --> 00:09:48,514 little bit and see what you think about 263 00:09:48,514 --> 00:09:50,639 it. To me, a lot of the problem 264 00:09:50,639 --> 00:09:53,120 with network engineering today is we can do 265 00:09:53,120 --> 00:09:54,980 everything 1,000 different ways. 266 00:09:55,759 --> 00:09:58,639 And none of them are, quote, unquote, wrong. 267 00:09:58,639 --> 00:10:00,800 I put the scare quotes around that, quote, 268 00:10:00,800 --> 00:10:01,779 unquote, wrong. 269 00:10:03,305 --> 00:10:04,045 I mean, 270 00:10:04,665 --> 00:10:05,805 how do you maintain 271 00:10:06,585 --> 00:10:07,325 a network 272 00:10:08,184 --> 00:10:10,745 when I don't know, I walk into networks, 273 00:10:10,745 --> 00:10:12,985 and they have five different tunneling protocols they're 274 00:10:12,985 --> 00:10:13,485 using, 275 00:10:13,865 --> 00:10:16,264 and five different ways of running BGP and 276 00:10:16,264 --> 00:10:16,925 five different 277 00:10:17,519 --> 00:10:18,019 IGPs. 278 00:10:18,799 --> 00:10:21,199 And, you know, that impacts maintainability. So how 279 00:10:21,199 --> 00:10:23,360 are y'all I mean, how do you think 280 00:10:23,360 --> 00:10:25,600 about that? You know, just as a as 281 00:10:25,600 --> 00:10:27,620 an engineer, how do you think about that? 282 00:10:28,000 --> 00:10:30,179 Yeah. I think you require sort of two 283 00:10:31,115 --> 00:10:33,754 seemingly dichotomous thoughts to be held together at 284 00:10:33,754 --> 00:10:34,575 the same time. 285 00:10:35,514 --> 00:10:38,014 One is you should be incredibly opinionated 286 00:10:39,034 --> 00:10:40,735 about how things should be implemented 287 00:10:41,914 --> 00:10:43,294 and go deep in that. 288 00:10:43,779 --> 00:10:46,100 But two, maybe this is, to what you 289 00:10:46,100 --> 00:10:47,940 were saying earlier, Russ, on how much the 290 00:10:47,940 --> 00:10:49,559 networking industry has changed. 291 00:10:50,820 --> 00:10:52,360 The industry has become 292 00:10:52,980 --> 00:10:53,480 buzzword 293 00:10:54,179 --> 00:10:55,240 alphabet soup 294 00:10:55,875 --> 00:10:57,014 acronym amalgamation 295 00:10:57,475 --> 00:10:58,855 the last two, three decades. 296 00:11:00,434 --> 00:11:03,075 As an engineer and also the ethos of 297 00:11:03,075 --> 00:11:04,754 meter is pretty simple on this, which is 298 00:11:04,754 --> 00:11:06,674 we just wanna do the basics of networking 299 00:11:06,674 --> 00:11:07,174 right. 300 00:11:07,794 --> 00:11:10,289 And because we're building the entire stack and 301 00:11:10,289 --> 00:11:12,769 not just like one small corner of it 302 00:11:12,769 --> 00:11:14,929 and then throwing it over the wall somebody 303 00:11:14,929 --> 00:11:16,629 else can maintain, we can do 304 00:11:17,250 --> 00:11:19,750 protocols really well, implementations of, 305 00:11:21,169 --> 00:11:23,750 implementations of these protocols really well and maintenance 306 00:11:23,809 --> 00:11:26,235 of these protocols. And then we're responsible for 307 00:11:26,235 --> 00:11:28,154 building the software all the way up to 308 00:11:28,154 --> 00:11:30,794 the APIs and applications where engineers and our 309 00:11:30,794 --> 00:11:32,254 customers will use them. 310 00:11:32,634 --> 00:11:34,735 So we do believe in being really opinionated 311 00:11:34,875 --> 00:11:36,174 on how to 312 00:11:36,730 --> 00:11:37,549 implement these 313 00:11:37,850 --> 00:11:40,570 things. But at the core of what meter 314 00:11:40,570 --> 00:11:43,049 is is just doing the basics in networking. 315 00:11:43,049 --> 00:11:43,549 Right. 316 00:11:44,330 --> 00:11:46,269 And I know that's might seem 317 00:11:47,610 --> 00:11:50,169 blase, but it doesn't seem like anybody else 318 00:11:50,169 --> 00:11:51,470 is doing it these days. 319 00:11:52,524 --> 00:11:54,705 Yeah. Interesting. Because, yeah, that's, 320 00:11:56,205 --> 00:11:58,125 that's another issue. And part of what you're 321 00:11:58,125 --> 00:12:01,424 also talking is like the whole idea that 322 00:12:01,884 --> 00:12:03,964 it's not just about what protocols you deploy. 323 00:12:03,964 --> 00:12:06,065 It's also about how those protocols are written. 324 00:12:06,519 --> 00:12:08,440 But yes, I still go back to my 325 00:12:08,440 --> 00:12:11,080 like, nerd knobs. Like, we gotta stop it 326 00:12:11,080 --> 00:12:13,259 with a nerd knobs, the nerd knobs are, 327 00:12:13,639 --> 00:12:14,700 are killing us. 328 00:12:15,080 --> 00:12:15,580 Unfortunately, 329 00:12:17,480 --> 00:12:19,240 I'm working in the lab right now working 330 00:12:19,240 --> 00:12:21,365 on something for not not for my job, 331 00:12:21,365 --> 00:12:23,605 but for something different. And I'm in this 332 00:12:23,605 --> 00:12:25,605 lab work, and I'm like, trying to figure 333 00:12:25,605 --> 00:12:28,004 out how to do something. And there's four 334 00:12:28,004 --> 00:12:29,625 apparent ways to do it. 335 00:12:30,325 --> 00:12:31,924 Yeah, but only one of them is going 336 00:12:31,924 --> 00:12:34,245 to work. Correct. How many hours did I 337 00:12:34,245 --> 00:12:36,105 go down trying to figure out like, 338 00:12:36,470 --> 00:12:38,230 okay, that looks like it should do what 339 00:12:38,230 --> 00:12:40,330 I think it should do. But it doesn't. 340 00:12:40,950 --> 00:12:42,870 So maybe that does something different. So then 341 00:12:42,870 --> 00:12:44,149 I gotta try to figure out what that 342 00:12:44,149 --> 00:12:46,389 does. I don't think it's the engineers fault, 343 00:12:46,389 --> 00:12:48,230 though. I think it's the vendors, because what 344 00:12:48,230 --> 00:12:50,174 was happening to all these vendors is everybody 345 00:12:50,414 --> 00:12:53,054 buys twenty, thirty companies a year, and those 346 00:12:53,054 --> 00:12:55,695 are twenty, thirty different teams that are building 347 00:12:55,695 --> 00:12:57,875 these features that are expected to work together. 348 00:12:58,254 --> 00:12:59,855 And then by the time the customers and 349 00:12:59,855 --> 00:13:01,695 engineers are using it, yeah, there are 20 350 00:13:01,695 --> 00:13:03,615 different ways of doing it. There's no one 351 00:13:03,615 --> 00:13:06,600 cohesive vision. It's not one cohesive product, not 352 00:13:06,600 --> 00:13:09,339 one cohesive API and not one cohesive application. 353 00:13:09,960 --> 00:13:12,039 This is like primarily what we disagree with, 354 00:13:12,039 --> 00:13:14,120 how the networking industry has happened. Nobody really 355 00:13:14,120 --> 00:13:16,039 builds stuff anymore. If you're a large vendor, 356 00:13:16,039 --> 00:13:17,799 you're using to try to buy companies left 357 00:13:17,799 --> 00:13:19,259 and right. And 358 00:13:19,644 --> 00:13:20,144 seemingly, 359 00:13:21,325 --> 00:13:23,804 this idea of I can buy two software 360 00:13:23,804 --> 00:13:25,664 and hardware companies and say, combine, 361 00:13:26,284 --> 00:13:28,524 that only happens in a business person's mind. 362 00:13:28,524 --> 00:13:30,144 That's not how engineering works. 363 00:13:30,605 --> 00:13:31,105 Yeah. 364 00:13:31,644 --> 00:13:34,544 And even and even to the same point, 365 00:13:35,179 --> 00:13:37,179 you're also faced with a situation when you 366 00:13:37,179 --> 00:13:39,339 get to a certain size, where you have 367 00:13:39,339 --> 00:13:41,500 customer a who wants to do x and 368 00:13:41,500 --> 00:13:43,519 customer b who wants to do y. 369 00:13:44,139 --> 00:13:45,899 And so when I worked at vendors at 370 00:13:45,899 --> 00:13:47,759 Cisco and Juniper, and even 371 00:13:49,105 --> 00:13:50,245 Erickson sausages, 372 00:13:50,545 --> 00:13:51,365 old redback, 373 00:13:51,745 --> 00:13:53,605 for those who know those those things. 374 00:13:55,745 --> 00:13:57,205 There was a lot of pressure 375 00:13:57,825 --> 00:13:58,565 to say, 376 00:13:59,184 --> 00:14:01,585 okay, customer a really solve needs to solve 377 00:14:01,585 --> 00:14:03,024 this problem, they would like to do it 378 00:14:03,024 --> 00:14:05,379 that way. Customer b would like to solve 379 00:14:05,379 --> 00:14:06,980 this problem, they'd like to solve it a 380 00:14:06,980 --> 00:14:09,000 different way, even if it's the same problem. 381 00:14:09,620 --> 00:14:12,019 These customers don't talk to each other. So 382 00:14:12,019 --> 00:14:13,940 they're pushing you for two different features to 383 00:14:13,940 --> 00:14:14,920 do the same thing. 384 00:14:15,460 --> 00:14:17,860 And you're ending up going, well, which one 385 00:14:17,860 --> 00:14:20,414 do I implement? Whoever's has the most money 386 00:14:20,414 --> 00:14:22,495 on the table first, if they both have 387 00:14:22,495 --> 00:14:24,334 a lot of money on the table, then 388 00:14:24,334 --> 00:14:25,875 you implement both of them. 389 00:14:27,054 --> 00:14:29,375 Even to the point where, when I was 390 00:14:29,375 --> 00:14:31,534 at one company, we had competing account teams 391 00:14:31,534 --> 00:14:33,235 going into a single account, 392 00:14:33,629 --> 00:14:36,350 and trying to sell them to different vendor 393 00:14:36,350 --> 00:14:36,850 products. 394 00:14:37,389 --> 00:14:39,389 That's right. It's only gotten worse since you've 395 00:14:39,389 --> 00:14:41,629 been away, by the way. It has gotten 396 00:14:41,629 --> 00:14:44,029 worse to the point where customers are like 397 00:14:44,029 --> 00:14:46,029 told not to buy the other thing by 398 00:14:46,029 --> 00:14:46,929 the same company. 399 00:14:47,790 --> 00:14:50,154 Yeah. Yeah. That's it is it's insane. 400 00:14:50,695 --> 00:14:52,475 And it does impact 401 00:14:53,095 --> 00:14:54,475 operations and serviceability. 402 00:14:55,735 --> 00:14:56,235 So 403 00:14:56,855 --> 00:14:58,554 let's go over a little bit of operations. 404 00:14:58,615 --> 00:15:01,495 So the first thing with operations is meantime 405 00:15:01,495 --> 00:15:02,394 between failure. 406 00:15:02,855 --> 00:15:05,149 That to me is the first thing. So 407 00:15:05,370 --> 00:15:07,289 there you're trying to address that, or you're 408 00:15:07,289 --> 00:15:10,169 saying we should address that by just reducing 409 00:15:10,169 --> 00:15:12,909 the diversity of operating systems stack software. 410 00:15:13,769 --> 00:15:14,269 And 411 00:15:14,809 --> 00:15:17,370 so that helps you to build higher quality 412 00:15:17,370 --> 00:15:17,870 software. 413 00:15:18,335 --> 00:15:20,754 Now there's a counter to that, which is, 414 00:15:21,294 --> 00:15:22,914 what do I do about monocultures? 415 00:15:23,934 --> 00:15:26,274 Because monocultures are a major problem. 416 00:15:27,774 --> 00:15:29,934 I remember once we were asked to design 417 00:15:29,934 --> 00:15:32,174 a network that would never fail by a 418 00:15:32,174 --> 00:15:32,914 big bank. 419 00:15:33,309 --> 00:15:34,830 We told them they couldn't afford it, and 420 00:15:34,830 --> 00:15:37,230 they said try me. So what we did 421 00:15:37,230 --> 00:15:39,230 was we built two network cores. It wasn't 422 00:15:39,230 --> 00:15:40,690 their whole network, just their core. 423 00:15:41,070 --> 00:15:42,610 We built parallel cores, 424 00:15:43,470 --> 00:15:45,710 one running one routing protocol, the other running 425 00:15:45,710 --> 00:15:48,110 another routing protocol, one running one piece of 426 00:15:48,110 --> 00:15:48,610 hardware 427 00:15:49,225 --> 00:15:51,304 with a one dot one operating network operating 428 00:15:51,304 --> 00:15:53,384 system, the other running a different piece of 429 00:15:53,384 --> 00:15:55,865 software, a different network network operating system. And 430 00:15:55,865 --> 00:15:57,945 it was specifically because of situation they'd hit 431 00:15:57,945 --> 00:15:59,085 was they had a monoculture, 432 00:15:59,705 --> 00:16:01,544 they got it, they caught a defect, but 433 00:16:01,544 --> 00:16:03,304 it was an all their package and all 434 00:16:03,304 --> 00:16:03,965 their hardware, 435 00:16:04,399 --> 00:16:05,460 and all their software. 436 00:16:05,840 --> 00:16:08,740 And that costs like, a massive outage. 437 00:16:09,120 --> 00:16:11,440 And they were very, very unhappy about the 438 00:16:11,440 --> 00:16:13,519 massive outage. And I don't blame them because 439 00:16:13,519 --> 00:16:14,899 it was on the network core. 440 00:16:16,160 --> 00:16:18,875 So what do you think about monocultures 441 00:16:19,254 --> 00:16:20,554 in relation to 442 00:16:21,095 --> 00:16:22,315 trying to unify 443 00:16:23,495 --> 00:16:26,455 all of this, everything into a single operating 444 00:16:26,455 --> 00:16:26,955 stack? 445 00:16:29,415 --> 00:16:31,495 Yeah. I mean, I think it depends on 446 00:16:31,495 --> 00:16:33,929 how it's built and why it's built. And 447 00:16:34,470 --> 00:16:35,929 is it built by, 448 00:16:36,950 --> 00:16:37,769 sort of 449 00:16:38,550 --> 00:16:40,550 trying to do some niche thing or working 450 00:16:40,550 --> 00:16:42,730 backwards from what customers are trying to do? 451 00:16:44,389 --> 00:16:46,324 We care a lot about mean time to 452 00:16:46,324 --> 00:16:46,824 resolution, 453 00:16:47,125 --> 00:16:49,365 not even mean time to failure. Like, how 454 00:16:49,365 --> 00:16:51,065 do we resolve these things? 455 00:16:52,644 --> 00:16:55,144 Then it's root cause for us on 456 00:16:55,845 --> 00:16:58,230 where did the issue happen. Right? So there 457 00:16:58,309 --> 00:17:01,449 are sort of three important bits in networking. 458 00:17:02,309 --> 00:17:04,169 One is network design. 459 00:17:05,109 --> 00:17:06,569 Second is network configuration, 460 00:17:06,869 --> 00:17:08,169 then it's network maintenance. 461 00:17:09,349 --> 00:17:11,269 So how do we get to mean time 462 00:17:11,269 --> 00:17:11,929 to resolution? 463 00:17:13,724 --> 00:17:16,845 Then understand where did the issue originate? Was 464 00:17:16,845 --> 00:17:18,365 it in the design of the network? Was 465 00:17:18,365 --> 00:17:20,285 it in the configuration of the network? Was 466 00:17:20,285 --> 00:17:22,625 it during maintenance when we're making changes because 467 00:17:22,924 --> 00:17:24,765 there's a bunch of new client devices that 468 00:17:24,765 --> 00:17:26,845 are being added, security rules, where did it 469 00:17:26,845 --> 00:17:27,345 happen? 470 00:17:28,589 --> 00:17:31,309 So if you work backwards from understanding what 471 00:17:31,309 --> 00:17:33,009 customer problems you're trying to solve 472 00:17:33,549 --> 00:17:34,930 and build applications 473 00:17:35,390 --> 00:17:37,549 like that, and then build APIs for those 474 00:17:37,549 --> 00:17:41,085 applications like that, then build operating systems like 475 00:17:41,085 --> 00:17:43,884 that, then build firmware like that, and build 476 00:17:43,884 --> 00:17:45,345 hardware to service that, 477 00:17:46,044 --> 00:17:48,125 then monocultures, I think, are less of an 478 00:17:48,125 --> 00:17:49,884 issue. They are certainly an issue with how 479 00:17:49,884 --> 00:17:51,585 the legacy companies do it. 480 00:17:53,450 --> 00:17:56,109 Yeah, interesting. So you're saying basically, 481 00:17:56,730 --> 00:17:58,990 the monoculture can be countered by 482 00:17:59,369 --> 00:18:02,349 focusing on quality and focusing on 483 00:18:02,650 --> 00:18:04,730 making sure you're doing the one thing you're 484 00:18:04,730 --> 00:18:05,230 after 485 00:18:05,924 --> 00:18:08,724 very well, instead of splitting yourself up and 486 00:18:08,724 --> 00:18:10,244 doing all sorts of trying to do all 487 00:18:10,244 --> 00:18:11,304 sorts of different things. 488 00:18:11,764 --> 00:18:12,264 Okay. 489 00:18:12,804 --> 00:18:14,644 So and you also said another interesting thing 490 00:18:14,644 --> 00:18:16,585 there, which is that you're focused on MTTR 491 00:18:17,125 --> 00:18:19,144 mean time to repair mean time to resolution. 492 00:18:20,190 --> 00:18:21,329 So that is 493 00:18:21,630 --> 00:18:22,130 another. 494 00:18:25,789 --> 00:18:28,429 That's another area that people often don't think 495 00:18:28,429 --> 00:18:30,450 about is that from an operational 496 00:18:30,750 --> 00:18:31,250 perspective, 497 00:18:32,804 --> 00:18:33,304 MTTR 498 00:18:33,684 --> 00:18:36,565 is actually far more important than MTBF. In 499 00:18:36,565 --> 00:18:38,265 most networks, most networks, 500 00:18:39,204 --> 00:18:40,424 failures are actually, 501 00:18:40,884 --> 00:18:42,644 I mean, I say, the law of large 502 00:18:42,644 --> 00:18:44,024 numbers kicks in, right? 503 00:18:44,339 --> 00:18:46,660 If I'm if I have, like, I used 504 00:18:46,660 --> 00:18:49,059 to work on data centers that were, I 505 00:18:49,059 --> 00:18:50,179 don't know, 10,000 506 00:18:50,660 --> 00:18:53,779 1,200 routers, 1,300 routers, and therefore, when you 507 00:18:53,779 --> 00:18:55,960 start thinking about 64 port routers 508 00:18:56,355 --> 00:18:58,434 or switches, and you start thinking about 64 509 00:18:58,434 --> 00:19:00,914 by 1,200, and you start thinking about optics 510 00:19:00,914 --> 00:19:01,414 and 511 00:19:02,115 --> 00:19:04,295 links and cables and 512 00:19:04,755 --> 00:19:06,835 power supplies and all the other odds and 513 00:19:06,835 --> 00:19:08,595 as they go into it, you can have 514 00:19:08,595 --> 00:19:11,174 really high MTVFs on each of those components, 515 00:19:11,720 --> 00:19:13,500 and you're still gonna have failures 516 00:19:13,880 --> 00:19:14,380 constantly. 517 00:19:14,839 --> 00:19:16,519 Like, two, three times a day, you're still 518 00:19:16,519 --> 00:19:18,600 gonna have failures. And that's just gonna be 519 00:19:18,680 --> 00:19:20,519 you have to work around it, in your 520 00:19:20,519 --> 00:19:21,420 network design. 521 00:19:22,200 --> 00:19:22,700 But 522 00:19:23,559 --> 00:19:25,555 beyond that, like, MTTR 523 00:19:26,015 --> 00:19:27,555 is very, very important. 524 00:19:28,255 --> 00:19:30,095 And I think you're right in that having 525 00:19:30,095 --> 00:19:32,194 a single user interface, a single 526 00:19:32,575 --> 00:19:34,414 way to get into things, and a single 527 00:19:34,414 --> 00:19:36,355 way things operate a single stack 528 00:19:36,734 --> 00:19:38,674 is probably a very big boon 529 00:19:39,119 --> 00:19:39,779 to MTTR 530 00:19:40,799 --> 00:19:42,720 count, you know, once you get past the 531 00:19:42,720 --> 00:19:43,220 monoculture 532 00:19:43,920 --> 00:19:44,420 issues. 533 00:19:45,839 --> 00:19:47,779 Now another part of that 534 00:19:49,279 --> 00:19:50,180 is that 535 00:19:51,599 --> 00:19:52,420 the interface. 536 00:19:55,275 --> 00:19:56,714 So a lot of the problem we face 537 00:19:56,714 --> 00:19:59,194 in the industry, and Greg Farrell was fussed 538 00:19:59,194 --> 00:20:01,115 about this, and Ethan's fussed about it, and 539 00:20:01,115 --> 00:20:03,295 I probably fussed about it at some point, 540 00:20:03,674 --> 00:20:05,855 is that all our interfaces are different. 541 00:20:07,549 --> 00:20:09,009 Like, s n m p MIB, 542 00:20:09,390 --> 00:20:11,730 like, whatever happened to that, like, that's gone. 543 00:20:12,430 --> 00:20:14,109 People still use it, but they only use 544 00:20:14,109 --> 00:20:15,089 it for some things. 545 00:20:15,549 --> 00:20:17,089 We're trying to move to Yang. 546 00:20:17,390 --> 00:20:20,029 But most people just use Ansible modules, the 547 00:20:20,029 --> 00:20:20,930 screen scrape. 548 00:20:21,554 --> 00:20:23,394 And like we're never really I don't think 549 00:20:23,394 --> 00:20:25,014 we're ever really going to get 550 00:20:25,315 --> 00:20:26,294 to an operationally 551 00:20:26,674 --> 00:20:28,375 secure and sound network, 552 00:20:28,835 --> 00:20:31,414 unless we make the machine interfaces work. 553 00:20:32,674 --> 00:20:34,914 It really worries me where we are with 554 00:20:34,914 --> 00:20:36,134 this, that we're still 555 00:20:36,839 --> 00:20:39,319 out here with Ansible modules that scrape five 556 00:20:39,319 --> 00:20:40,619 different kinds of routers. 557 00:20:41,640 --> 00:20:42,140 And 558 00:20:42,519 --> 00:20:45,259 even for Yang, we're doing Yang conversion modules 559 00:20:45,319 --> 00:20:45,980 and stuff. 560 00:20:46,279 --> 00:20:47,980 And that that actually really 561 00:20:48,599 --> 00:20:51,500 disturbs me. So any thoughts on that realm 562 00:20:51,799 --> 00:20:52,299 of 563 00:20:52,985 --> 00:20:54,045 how things go? 564 00:20:56,184 --> 00:20:57,705 Yeah. It comes back to, again, what we 565 00:20:57,705 --> 00:20:59,325 spoke about earlier, which is 566 00:20:59,705 --> 00:21:01,164 trying to solve the problem 567 00:21:02,744 --> 00:21:04,985 from a customer perspective, and then working backwards 568 00:21:04,985 --> 00:21:05,644 from that. 569 00:21:06,559 --> 00:21:08,640 In this case, what are customers trying to 570 00:21:08,640 --> 00:21:11,039 do? Whether it's with Ansible or Chef or 571 00:21:11,039 --> 00:21:12,259 Yang or Terraform, 572 00:21:13,119 --> 00:21:13,619 anything, 573 00:21:14,480 --> 00:21:16,420 is just to be able to do something 574 00:21:17,039 --> 00:21:17,539 repeated 575 00:21:18,825 --> 00:21:21,625 And then without having to know every single 576 00:21:21,625 --> 00:21:24,024 minutia of commands and remembering that for every 577 00:21:24,024 --> 00:21:25,865 single piece of hardware and every single route 578 00:21:25,865 --> 00:21:27,005 and every single protocol, 579 00:21:27,544 --> 00:21:29,065 that's what they're trying to get to. 580 00:21:29,784 --> 00:21:31,644 We obviously started with CLI 581 00:21:32,265 --> 00:21:32,765 in 582 00:21:33,130 --> 00:21:33,630 networking, 583 00:21:34,089 --> 00:21:37,470 then 20 ago sort of moved to dashboards. 584 00:21:38,490 --> 00:21:40,809 We released a product called Command about a 585 00:21:40,809 --> 00:21:42,409 year, year and a half ago now that's 586 00:21:42,409 --> 00:21:45,230 gotten really great responses from customers and partners 587 00:21:45,929 --> 00:21:46,424 that addresses 588 00:21:47,545 --> 00:21:48,684 some of these issues of 589 00:21:49,465 --> 00:21:51,465 being able to not needing to know every 590 00:21:51,465 --> 00:21:51,965 single, 591 00:21:52,825 --> 00:21:54,745 you know, little detail on a command and 592 00:21:54,745 --> 00:21:56,505 things, but really being able to do automation 593 00:21:56,505 --> 00:21:58,825 at scale and really being able to build 594 00:21:58,825 --> 00:22:00,825 networks at scale. So I do think it's 595 00:22:00,825 --> 00:22:01,644 really important. 596 00:22:02,690 --> 00:22:05,009 I agree and I understand where your worries 597 00:22:05,009 --> 00:22:07,570 coming from. But again, working backwards from what 598 00:22:07,570 --> 00:22:09,570 customers are trying to do, and our view 599 00:22:09,570 --> 00:22:10,470 is, you know, 600 00:22:11,089 --> 00:22:12,230 being able to build 601 00:22:12,609 --> 00:22:15,035 networks repeatedly of the same quality and of 602 00:22:15,035 --> 00:22:15,694 the same 603 00:22:15,994 --> 00:22:16,494 configuration, 604 00:22:17,275 --> 00:22:19,434 and having to do it without having to 605 00:22:19,434 --> 00:22:19,934 remember, 606 00:22:20,394 --> 00:22:23,694 you know, every single small command line 607 00:22:24,555 --> 00:22:25,934 incantation as possible. 608 00:22:26,920 --> 00:22:28,140 Interesting. Okay. 609 00:22:29,080 --> 00:22:29,580 So 610 00:22:30,039 --> 00:22:32,359 another realm in which I think about operations 611 00:22:32,359 --> 00:22:33,720 a lot that we don't address a lot 612 00:22:33,720 --> 00:22:36,380 in the networking community or network engineering world 613 00:22:36,680 --> 00:22:37,820 is just simply 614 00:22:38,519 --> 00:22:39,019 testing. 615 00:22:39,720 --> 00:22:42,220 And I don't just mean unit testing, 616 00:22:42,625 --> 00:22:44,224 because of course, when I work for a 617 00:22:44,224 --> 00:22:44,724 vendor, 618 00:22:45,025 --> 00:22:47,265 and even now, I still deal with unit 619 00:22:47,265 --> 00:22:50,484 testing versus system testing, and the combinatorial explosion 620 00:22:51,025 --> 00:22:53,525 of system testing across multiple units. 621 00:22:54,144 --> 00:22:55,125 So is that 622 00:22:56,859 --> 00:22:59,679 something that you think about from an operational 623 00:22:59,740 --> 00:23:01,119 perspective? Is this like, 624 00:23:01,740 --> 00:23:03,919 again, just as an engineer, do you like, 625 00:23:04,380 --> 00:23:06,700 think that this is an issue? And, you 626 00:23:06,700 --> 00:23:09,500 know, is there solutions you think are coming 627 00:23:09,500 --> 00:23:10,559 out there or 628 00:23:10,875 --> 00:23:11,615 anything like 629 00:23:12,634 --> 00:23:14,795 that? Yes. I do think it's an issue, 630 00:23:14,795 --> 00:23:15,855 but I'm pretty sure, 631 00:23:16,315 --> 00:23:18,234 there's not an engineer that's serious about their 632 00:23:18,234 --> 00:23:20,015 work that does not think it's an issue. 633 00:23:22,990 --> 00:23:25,950 But maybe you you'll hear me as a 634 00:23:25,950 --> 00:23:26,929 broken record, 635 00:23:27,630 --> 00:23:30,210 but we we do think here the integration 636 00:23:30,349 --> 00:23:32,849 and the thoughtfulness from the ground up matters. 637 00:23:33,630 --> 00:23:36,029 Here's how we do it, and maybe to 638 00:23:36,029 --> 00:23:37,005 give you an example, 639 00:23:37,384 --> 00:23:38,765 I know this is a, 640 00:23:39,304 --> 00:23:41,944 this is sort of a audio only version 641 00:23:41,944 --> 00:23:43,625 of of this of this talk, but, you 642 00:23:43,625 --> 00:23:46,585 know, happy to show Yeah. Specific instances at 643 00:23:46,585 --> 00:23:47,884 some point. But 644 00:23:48,920 --> 00:23:51,240 if you think about how networking started, most 645 00:23:51,240 --> 00:23:53,579 of the intelligence on how to do routing, 646 00:23:53,640 --> 00:23:56,039 what the configuration is, how to test the 647 00:23:56,039 --> 00:23:59,160 network was basically all inside the hardware as 648 00:23:59,160 --> 00:24:00,059 much as possible. 649 00:24:01,075 --> 00:24:03,475 Then obviously, the buzziest thing that happened in 650 00:24:03,475 --> 00:24:05,795 networking twenty years ago or fifteen years ago 651 00:24:05,795 --> 00:24:07,095 was software defined networking. 652 00:24:07,634 --> 00:24:08,134 And 653 00:24:08,434 --> 00:24:10,914 the idea was to move a lot of 654 00:24:10,914 --> 00:24:13,609 the definition of this networking into software 655 00:24:14,230 --> 00:24:15,369 so that it's malleable 656 00:24:15,670 --> 00:24:16,490 to do it. 657 00:24:17,190 --> 00:24:18,549 What I think we've done over the last 658 00:24:18,549 --> 00:24:20,630 five, seven so five or seven years is 659 00:24:20,630 --> 00:24:21,450 a little bit 660 00:24:21,990 --> 00:24:24,569 interesting and sort of next step after that 661 00:24:24,789 --> 00:24:25,930 is we moved 662 00:24:26,390 --> 00:24:27,529 all of the intelligence 663 00:24:31,515 --> 00:24:33,275 of the of the network actually into our 664 00:24:33,275 --> 00:24:35,695 back end down to a single port. 665 00:24:36,234 --> 00:24:38,494 What that means is we can actually configure 666 00:24:39,275 --> 00:24:41,730 thousands of pieces of hardware down to the 667 00:24:41,730 --> 00:24:44,309 protocol level, how they're connected, how 668 00:24:44,610 --> 00:24:45,910 the topology is, 669 00:24:46,289 --> 00:24:48,930 make sure the configuration is right on these 670 00:24:48,930 --> 00:24:49,430 networks. 671 00:24:49,970 --> 00:24:50,789 And only 672 00:24:51,170 --> 00:24:52,930 when we are happy with it and a 673 00:24:52,930 --> 00:24:54,470 physical hardware is connected, 674 00:24:56,085 --> 00:24:58,565 the hardware downloads the source of truth from 675 00:24:58,565 --> 00:24:59,924 the back end, and then it's off to 676 00:24:59,924 --> 00:25:01,765 the races. But what that enables us to 677 00:25:01,765 --> 00:25:02,265 do 678 00:25:02,884 --> 00:25:05,284 is to be able to configure and test 679 00:25:05,284 --> 00:25:06,025 these networks 680 00:25:07,765 --> 00:25:08,984 virtually first. 681 00:25:10,109 --> 00:25:12,369 And then when the hardware is plugged in, 682 00:25:12,589 --> 00:25:14,750 these virtual networks are the source of truth 683 00:25:14,750 --> 00:25:16,450 that come down into the hardware. 684 00:25:16,750 --> 00:25:19,069 So it's a remarkable system that's enabled us 685 00:25:19,069 --> 00:25:21,470 to scale a lot, and it's one of 686 00:25:21,470 --> 00:25:22,529 our sort of 687 00:25:22,990 --> 00:25:23,490 differentiating 688 00:25:23,789 --> 00:25:26,744 ways to make sure networks are configured properly 689 00:25:26,964 --> 00:25:29,285 because configurations are where majority of the issues 690 00:25:29,285 --> 00:25:31,444 happen and then being able to test these 691 00:25:31,444 --> 00:25:33,605 networks and then maintain these networks. So we 692 00:25:33,605 --> 00:25:34,825 moved everything 693 00:25:35,605 --> 00:25:38,325 into a virtual network which is how all 694 00:25:38,325 --> 00:25:40,025 of our networks are first configured. 695 00:25:41,170 --> 00:25:43,730 Then when the hardware is plugged in, it's 696 00:25:43,730 --> 00:25:46,210 simply a mirroring of that virtual network. So 697 00:25:46,210 --> 00:25:47,910 that's been our solution for 698 00:25:49,170 --> 00:25:50,230 it. Okay. Interesting. 699 00:25:50,529 --> 00:25:52,609 So but in the larger network world, you 700 00:25:52,609 --> 00:25:54,369 would say that it's still a concern. People 701 00:25:54,369 --> 00:25:56,150 are still very worried about this. 702 00:25:56,625 --> 00:25:57,125 And, 703 00:25:57,424 --> 00:25:59,445 yeah, and it's still an issue 704 00:25:59,825 --> 00:26:01,505 that needs to be addressed in some way. 705 00:26:01,505 --> 00:26:03,525 Now another thing that I often think about 706 00:26:03,904 --> 00:26:05,605 is in terms of operations 707 00:26:06,065 --> 00:26:07,125 is lifecycle. 708 00:26:07,744 --> 00:26:10,224 Another thing that we pay zero attention to 709 00:26:10,224 --> 00:26:12,269 in the network world, I have 710 00:26:12,809 --> 00:26:14,589 seen routers handed down 711 00:26:15,289 --> 00:26:17,789 from place to place inside an organization 712 00:26:18,569 --> 00:26:21,130 and use for five different roles. Because once 713 00:26:21,130 --> 00:26:22,970 you buy it, you own it. Now I 714 00:26:22,970 --> 00:26:24,650 know your business model is a little bit 715 00:26:24,650 --> 00:26:26,970 different than that. But do you think that 716 00:26:26,970 --> 00:26:27,470 there's 717 00:26:28,144 --> 00:26:30,625 what would your advice be to a network 718 00:26:30,625 --> 00:26:31,125 engineer 719 00:26:31,744 --> 00:26:32,804 beyond just, 720 00:26:33,265 --> 00:26:33,845 you know, 721 00:26:34,464 --> 00:26:36,544 switch to our model? Is there any other 722 00:26:36,544 --> 00:26:39,024 advice you could give network engineers about that 723 00:26:39,024 --> 00:26:40,005 problem? Because 724 00:26:40,339 --> 00:26:41,539 I don't I don't know how to fix 725 00:26:41,539 --> 00:26:43,619 it personally other than go to my manager 726 00:26:43,619 --> 00:26:44,359 and say, 727 00:26:45,619 --> 00:26:47,000 just buy new stuff. 728 00:26:47,380 --> 00:26:48,200 Stop it. 729 00:26:49,859 --> 00:26:53,220 I think, network engineers can make finance teams 730 00:26:53,220 --> 00:26:55,160 and procurement teams their best friends. 731 00:26:55,744 --> 00:26:57,205 And then that's how it can happen. 732 00:26:57,744 --> 00:26:59,904 Or switch to METERS model. Sorry. I had 733 00:26:59,904 --> 00:27:01,285 to say it. Yeah. That's fine. 734 00:27:01,664 --> 00:27:03,424 So when you talk about making that, you 735 00:27:03,424 --> 00:27:05,365 mean, like, go play golf with them. 736 00:27:06,225 --> 00:27:07,984 Yeah. Take them out to dinner, make sure 737 00:27:07,984 --> 00:27:10,789 they know how important the network is. Yeah. 738 00:27:11,029 --> 00:27:13,110 Make sure they understand that everything runs on 739 00:27:13,110 --> 00:27:13,690 the network. 740 00:27:14,230 --> 00:27:15,990 And if the network isn't great or the 741 00:27:15,990 --> 00:27:17,130 network isn't secure, 742 00:27:17,830 --> 00:27:20,170 how detrimental it is for the entire organization. 743 00:27:20,470 --> 00:27:21,910 Because I think a lot of times these 744 00:27:21,910 --> 00:27:22,730 days especially, 745 00:27:23,924 --> 00:27:26,105 people don't realize the importance of networking, 746 00:27:27,125 --> 00:27:29,284 and it's sort of an afterthought. So I 747 00:27:29,284 --> 00:27:30,644 do think there is sort of a shift 748 00:27:30,644 --> 00:27:32,644 in the network engineer's job of making sure 749 00:27:32,644 --> 00:27:33,784 people actually understand. 750 00:27:34,404 --> 00:27:35,944 It used to be that everybody 751 00:27:37,710 --> 00:27:39,650 knew the network is the heart and 752 00:27:40,109 --> 00:27:40,590 soul of it. But I don't think that's 753 00:27:40,590 --> 00:27:41,609 true anymore. And I think it's the job 754 00:27:41,609 --> 00:27:43,329 of the network engineer to make sure everybody 755 00:27:43,710 --> 00:27:44,210 understands. 756 00:27:46,190 --> 00:27:48,210 Yeah. And I think that's part of it 757 00:27:48,269 --> 00:27:48,930 is that 758 00:27:49,944 --> 00:27:52,424 we actually play against ourselves when we call 759 00:27:52,424 --> 00:27:53,724 the network a commodity. 760 00:27:55,544 --> 00:27:57,244 That that actually hurts us. 761 00:27:57,704 --> 00:27:58,204 Because 762 00:27:58,585 --> 00:28:00,024 people say, well, if it's a net, if 763 00:28:00,024 --> 00:28:00,764 it's a commodity, 764 00:28:01,079 --> 00:28:02,599 I just want you to add more bandwidth. 765 00:28:02,599 --> 00:28:04,519 It's like adding memory, right? You just add 766 00:28:04,519 --> 00:28:06,279 more bandwidth, like it's, it's not a bit 767 00:28:06,279 --> 00:28:07,339 just add bandwidth, 768 00:28:07,799 --> 00:28:09,880 you know, or just just make it faster, 769 00:28:09,880 --> 00:28:11,720 like it's a commodity, I can make a 770 00:28:11,720 --> 00:28:13,799 car that goes faster. So just make it 771 00:28:13,799 --> 00:28:14,539 go faster. 772 00:28:14,944 --> 00:28:17,365 Like, it's actually not like that. 773 00:28:18,304 --> 00:28:20,565 Routers and switches might be commodities. 774 00:28:20,944 --> 00:28:23,845 Fine. Got it. They're cattle, not pets. 775 00:28:24,464 --> 00:28:25,204 The network, 776 00:28:25,505 --> 00:28:27,924 in my opinion, is still not a commodity. 777 00:28:29,080 --> 00:28:31,400 The network is a system that is built 778 00:28:31,400 --> 00:28:32,860 up of commodity pieces. 779 00:28:33,240 --> 00:28:35,660 But that doesn't mean the network itself 780 00:28:35,960 --> 00:28:36,860 is a commodity. 781 00:28:37,480 --> 00:28:39,880 So I think we confuse this, and we 782 00:28:39,880 --> 00:28:42,039 work against ourselves, and we say these things, 783 00:28:42,039 --> 00:28:43,900 and then we need to realize, you know, 784 00:28:44,284 --> 00:28:46,384 we need to change our language. 785 00:28:47,164 --> 00:28:48,625 And the way we approach 786 00:28:49,244 --> 00:28:52,784 the company management, corporate management, and financial people, 787 00:28:53,005 --> 00:28:55,565 to help them understand that you can buy 788 00:28:55,565 --> 00:28:58,839 commodity hardware, fine, whatever, don't care. But your 789 00:28:58,839 --> 00:29:00,220 network is not a commodity. 790 00:29:00,679 --> 00:29:02,200 You need to take care of the network 791 00:29:02,200 --> 00:29:04,039 as a system. And that's still going to 792 00:29:04,039 --> 00:29:06,440 cost money. So that's, that's kind of a 793 00:29:06,440 --> 00:29:08,859 common approach for me, at any rate, 794 00:29:09,319 --> 00:29:10,779 is to do it that way. 795 00:29:11,480 --> 00:29:12,779 Yeah, make sense. 796 00:29:13,894 --> 00:29:14,394 So 797 00:29:15,494 --> 00:29:17,015 I don't have a whole lot more topics 798 00:29:17,015 --> 00:29:18,054 to talk about. I mean, I have a 799 00:29:18,054 --> 00:29:19,755 couple but I don't want to dive into 800 00:29:20,214 --> 00:29:22,394 yet another half an hour's worth of discussion. 801 00:29:23,255 --> 00:29:25,194 So we should just wrap it up there. 802 00:29:25,829 --> 00:29:27,829 Anything you want to say before we jump 803 00:29:27,829 --> 00:29:28,490 off, Anil? 804 00:29:29,509 --> 00:29:32,490 Networking is important. More people should study it. 805 00:29:32,789 --> 00:29:34,329 Networking should be cool again. 806 00:29:34,789 --> 00:29:37,269 Yes, I agree. Yeah, we're definitely in this 807 00:29:37,269 --> 00:29:38,089 mode where 808 00:29:38,975 --> 00:29:40,815 we don't have enough network engineers coming into 809 00:29:40,815 --> 00:29:41,955 the into the market. 810 00:29:43,215 --> 00:29:44,894 I look, I've been doing this for thirty 811 00:29:44,894 --> 00:29:47,455 years, I suppose, something like that. I look 812 00:29:47,455 --> 00:29:49,695 behind me. And not only do I not 813 00:29:49,695 --> 00:29:51,055 see a lot of people coming in, oh, 814 00:29:51,055 --> 00:29:53,250 that's not absolutely true. You know, I teach 815 00:29:53,250 --> 00:29:55,210 a couple of classes in network engineering, and 816 00:29:55,210 --> 00:29:57,230 I know people other people who teach classes. 817 00:29:57,529 --> 00:29:59,470 So I know there are people out there. 818 00:29:59,529 --> 00:30:01,369 I don't know if there's the quantity that 819 00:30:01,369 --> 00:30:02,029 we need. 820 00:30:02,330 --> 00:30:05,049 And to be honest, not to throw shade 821 00:30:05,049 --> 00:30:05,710 on people, 822 00:30:06,085 --> 00:30:08,244 I think the way we teach networking now 823 00:30:08,244 --> 00:30:10,585 actually develops people in the wrong way. 824 00:30:11,205 --> 00:30:11,705 Like, 825 00:30:13,365 --> 00:30:15,045 I'm actually write a post about this at 826 00:30:15,045 --> 00:30:17,924 some point about teaching from the outside instead 827 00:30:17,924 --> 00:30:20,164 of the inside out, and how we teach 828 00:30:20,164 --> 00:30:22,009 from the outside in, and how much I 829 00:30:22,009 --> 00:30:23,069 think that's a problem. 830 00:30:23,609 --> 00:30:26,029 So, yeah. Alright. Cool. So, 831 00:30:26,970 --> 00:30:28,329 Neil, where can people get in touch with 832 00:30:28,329 --> 00:30:29,450 you if they want to? Do you have 833 00:30:29,450 --> 00:30:30,669 a blog? Do you 834 00:30:31,289 --> 00:30:33,434 LinkedIn? Do you what do you do? 835 00:30:34,315 --> 00:30:35,215 Sure. LinkedIn. 836 00:30:36,394 --> 00:30:38,075 Our website is meter.com, 837 00:30:38,075 --> 00:30:39,355 metr.com. 838 00:30:39,355 --> 00:30:42,715 My email is pretty simple. Anil@meter.com. 839 00:30:42,715 --> 00:30:45,115 People can just email me. Okay. And but 840 00:30:45,115 --> 00:30:46,555 you don't have a blog or anything like 841 00:30:46,555 --> 00:30:48,730 that. You don't do any writing. No. Alright. 842 00:30:48,730 --> 00:30:50,910 Just we just write on the meter blog. 843 00:30:50,970 --> 00:30:53,210 Okay. Cool. Alright. And I'm Russ White. You 844 00:30:53,210 --> 00:30:54,410 can always find me here at the hedge. 845 00:30:54,410 --> 00:30:55,769 You can find me at rule eleven dot 846 00:30:55,769 --> 00:30:56,269 tech. 847 00:30:56,809 --> 00:30:58,250 You can find me on LinkedIn. You can 848 00:30:58,250 --> 00:31:01,049 find me on x, me, Twitter, whatever you 849 00:31:01,049 --> 00:31:03,194 want to call that thing. I'm actually really 850 00:31:03,194 --> 00:31:05,115 easy to find. If you have topic suggestions 851 00:31:05,115 --> 00:31:06,714 for the hedge, please let me know or 852 00:31:06,714 --> 00:31:08,794 people you'd like to see on. And we 853 00:31:08,794 --> 00:31:10,894 know we live in an attention driven economy. 854 00:31:11,355 --> 00:31:13,275 So you spending the time to listen to 855 00:31:13,275 --> 00:31:14,910 us all the way to the bitter end 856 00:31:15,150 --> 00:31:16,750 is really important and we thank you for 857 00:31:16,750 --> 00:31:18,430 spending the time with us and we will 858 00:31:18,430 --> 00:31:19,570 catch you next time.