Mohit Tahiliani 0:00 We were at the hackathon this week. A lot of people came to our table where we were doing the hackathon, and they were pretty appreciative about the effort that the students were putting in. And in fact, during that time, they offered two slots to the students to come and talk in their working group. So what the chairs from CCWG came and said, "Would you please present this what you have done in one of our meetings", it was very kind of them to give us the opportunity, and students were really thrilled, and they made the most out of it, we also got NMOP chairs telling us that if you could present this, it will be great. So students, again, took that opportunity, and I'm very grateful to them to give us these chances. George Michaelson 0:47 You're listening to ping, a podcast by APNIC, discussing all things related to measuring the internet. I'm your host. George Michaelson, welcome back for season six. We start the year with a set of three conversations held at IETF Bangkok last year. Two are with groups of students, and one with their supervising professor. The students are all from NITK, Karnataka, a technically focused university in the Indian state of Karnataka down south. It's located on the coast close to Mangalore. NITK has established a track record of network related research, encouraging undergraduate and post graduate students from their computer science and network engineering schools to come into the IETF, work at the hackathon and present their ideas in the appropriate working groups. We've heard from NITK students before on PING two attendees at IETF Brisbane came on the podcast and have subsequently moved into industry and postdoc research. They're developing their careers. We talk to them about their work on a campus IPV6 deployment, a multi year activity to deploy campus wide IPV6 that's been running at NITK. The project encourages students to mentor successive cohorts as they build out systems. This time, the two groups of students came to IETF and worked on different problems. One set looked at variants of the FQ-CODEL queueing disciplines, fair or flow queue management with control delay. The other team worked on the YANG data model in the NETCOM area. I also talked to their supervisor, Professor Mohit Tahliani. Mohit is working to solve the IETF's age problem, as well as to improve his students chances in the job market. He's hopeful some will stay around for the longer term in network standards development at the IETF, first up, let's hear about FQ-Codel Preeti. Can you explain to people who you are and what you're doing? Rati Preethi Subramanian 2:58 Hello, George. I'm Rathi Preethi, and I'm an undergrad student from National Institute of Technology Karnataka, and currently I'm pursuing my major in computer science, and I've been introduced to networking by Professor Mohit, and I've been working with him for like, one year now, and we have been working on projects related to multi path QUIC and congestion control. Shriya Anil 3:24 I'm Shriya Anil. I'm also studying in the National Institute of Technology Karnataka, and my major is CS as well. Mahati Kalale 3:30 Hello. I'm Mahati Kalale. I'm also a final year B Tech student studying computer science at National Institute of Technology Karnataka. Anuhya Murki 3:38 Hello. I'm Anuhya Murki. I am also doing a major in computer science of engineering and National Institute of Technology Karnataka. Supradha Bhat 3:47 Hello. I'm Supradha Bhat, Institute of Technology Karnataka, majoring in CS. We are excited to be here. George Michaelson 3:53 So this isn't the first time we've had people from NITK on PING.. Last year, we were speaking with Kavya and Vanessa, who were working on a campus IPV6 project. But you guys, you're actually here doing something quite different, aren't you? So you're working with Professor Mohit Tahiliani, who's the head of the Central computing facility at NITK, but he's also an academic, and he's been teaching on the course right? NITK, the National Institute for Technology in Karnataka. Can one of you just explain to people where Karnataka is? Anuhya Murki 4:30 So, Karnataka is a state in India on the South West, and Surathkal is a city in Karnataka [George: right] And it's right by the coast. So we have like a beach in our college, George Michaelson 4:42 and it's quite a large university. There's about 7 - 8000 students on campus. Mahati Kalale 4:48 That's right, the campus area is around 200 acres, and more than 7000 students pursuing various programs in the campus. George Michaelson 4:57 So you came to IETF and this time, you were doing something different. Now you mentioned that you were looking at some aspects of networking. This is in the space of queuing systems, isn't it, queuing disciplines. Could you talk a little bit about the ideas here? Rati Preethi Subramanian 5:15 So our professor has recently proposed an internet draft on FQ_pie, which is a Queue management discipline. And we also have a PhD student who is working on an update to FQ_Codel, which is called FQ_Codel++. So the main aim of our hackathon is to like test different queue disciplines using a network testing tool called CCPerf George Michaelson 5:39 Right? so queuing disciplines, they're bits of code that go inside the network system, in an operating system, they might be in the kernel, or they might be in user space, and users get to choose these is there like a tuning mechanism to select one? Mahati Kalale 5:55 Right So these queuing disciplines are mostly deployed at the router, and yes, you a user, can choose the queuing discipline that they want to deploy in their system George Michaelson 6:04 And the FQ part, I'm right in thinking that stands for fair queuing. The idea here is to distribute Mahati Kalale 6:11 both flow and fair queue George Michaelson 6:13 flow and fair queuing. How did you go about doing this? You were using a testing framework. I think Supradha Bhat 6:21 so we were using a testing framework called CCPerf, which was having NS3 as a back end. The algorithms FQ_pie and FQ_codel and many others are implemented in that. And we were testing out the implementations of FQ_pie, and our main aim was to compare it with FQ_codel. So that was basically our aim at the hackathon, and we did observe some interesting results which we had presented. George Michaelson 6:48 Anuhya, can you tell me a little bit about how you structured the work? Because the hackathon is really only one or two days of time to work in. So how did you make this work? Anuhya Murki 6:58 Yeah, so basically, all of us got familiar with the framework CCperf. We got ready all the installations and everything on our laptops, and so CCperf provides us with a list of experiments, and there are multiple categories under which we can work. Like, for example, we have a competition, we have network variation, etc. So we are a team of five members, and we had divided the experiments among ourselves. So each one of us worked on our own. George Michaelson 7:27 So you all installed NS3 on your laptops or on one machine that you had shared access to, Anuhya Murki 7:33 no our own personal systems, we had installed NS3 and CCperf, and we continued with our experiments. George Michaelson 7:42 Okay, so Mahati, how did you do the analysis of the data that you were getting? Did you construct a like a Jupyter working book or a spreadsheet? How were you doing a comparison? Mahati Kalale 7:54 So CCPerf provides us with statistics after we run a particular Q discipline, so we get to analyze throughput, inside-packet, SRTT, etc, and we we plotted graphs, and we mainly compare graphs to compare different Q disciplines and get results and analyze it. George Michaelson 8:14 Shriya, you presented on this in the hackathon, but also in the CCWG. I believe [Shriya yes, yes we did] So this work is something that might carry on. It kind of gives them a sense of these different forms of queuing, and they're thinking about progress here. Shriya Anil 8:32 Yeah. So our professor is planning to make a FQ_pie, a new internet, a new RFC for IETF, and these graphs will help him like so yeah, we also presented our findings in the CCWG working group. The graphs he produced will be used by a professor to help validate his findings for FQ_pie, which he plans to turn into an ID for it, an RFC for IETF. The graphs can also be used to help compare codel and codal++ Supradha Bhat 9:03 And one thing I would add on is that, basically, we wanted to introduce CCperf as a new tool, which can be, in general, used by the community to test out algorithms. And also we presented one graph, one plot, of BBR in the CCWG meeting. So. You know, we wanted to just spread awareness that these tools exist and might be useful. People can try it out George Michaelson 9:26 so Preethi. What kind of things do you think might follow on from this? Are there other projects that other NITK students might do next year or in another IETF? Rati Preethi Subramanian 9:36 So during our presentation in CCWG, we have presented that we have compared the CCperf experiments with the evaluation criteria that is specified in RFC 9743 we have found out that quite a lot of experiments that are listed in the CCperf tool are in alignment with the RFC, but there are certain evaluation criteria that is mentioned in the RFC that is not already there in CCperf. So as a continuation to our work, we can try to add those evaluation criteria or like experiments related to those evaluation that are not there in CCperf all day. George Michaelson 10:13 Oh, that's really great. Where do you see your future in this? Are you guys staying in the sector, or are you going out into industry? Is anyone going on to higher education? Shriya Anil 10:24 Yes, I'll be going for higher education in the US. I've applied for a few colleges and got some answers, George Michaelson 10:29 and this is to do work in networking. Shriya Anil 10:32 Yeah, yeah, I focus on networking in my SOP and the rest of you. Are you going into industry? Do you have jobs lined up? Mahati Kalale 10:40 Yes, the rest of us have all offers from companies. We were not yet sure which team we will be working with, but hopefully networking it would be great for us, because we have a strong background George Michaelson 10:52 this kind of experiment. It was fun to do. Anuhya Murki 10:55 Yes, it was really fun. So we had to learn about how new tools work. We got to see the behavior of different queue discipline algorithms. And it was very fascinating when what we learned theoretically is seen in the experiment results like, what we study is like, Okay, this is right, what we learned about this algorithm is right, or if it's wrong, we try to analyze why exactly it's going wrong and what parameter has to be tuned or what has to be changed in the algorithm. Yeah. It was very new and interesting to work on George Michaelson 11:24 that's really great. Thanks guys, thanks for coming on, PING. Mahati Kalale 11:27 Thank you so much. We all had a great time talking to you. Thank you. Thanks a lot, George, thank you. George Michaelson 11:34 Now let's hear from the NITK students about YANG data models in NETCONF. Vartika, Hayyan. Sid, Meher, welcome to PING. Can you introduce yourself and tell the listeners a little bit about who you are and why you're here? Vartika T Rao 11:53 Hello. I'm Vartika T Rao. I'm a fourth year undergrad student at NITK Surathkal. I am currently majoring in information technology. I also have a minor in machine learning. Networking is a little new to me. I've become more interested in it in the past year, and this has been a great opportunity to learn more about it. And I'm glad I'm here. George Michaelson 12:14 We all have to start somewhere. Hayyan Hayyan Arshad 12:17 Hello, I'm Hayyan, and I am a fourth year undergraduate from NITK Surathkal, and I became interested in networking through the courses I've taken under my professor Mohit Tahiliani George Michaelson 12:29 Sid Siddharth Bhat 12:29 Hi everyone. I'm Siddharth. So I'm also fourth year undergraduate at NITK Surathkal in Mangalore . I'm an electrical and electronics major, and I've also minor in machine learning. I would not say networking is my strength, but it's definitely extremely interesting, and I had a lot of fun in the whole process. George Michaelson 12:49 Meher Bharadwaja Meherrushi Chittapragada 12:49 my name is Bharadwaja Meherrushi. I'm also a fourth year computer science student at National Institute of Technology Karnataka. I have a minor in management, but I'm interested in most of the computer science fields, including systems. George Michaelson 13:04 So the four of you are here on a project we're going to talk about in a little while, under the supervision of Professor Mohit, who's the director of the computing center in the institute in Karnataka. But this isn't a project that started here, is it? I think you said Meher that you were working on this at a previous IETF in Dublin.? Bharadwaja Meherrushi Chittapragada 13:22 Yes, that is right. So I've been interested in EBPF as well, and few of my seniors were sort of working on an EBPF project, and they did present it in Brisbane. And that's how I went to IETF 121 where I also met Mahesh and Kent who were draft authors. So my professor actually talked to them, and they asked for students would be interested in implementing their draft called HTTPS NOTIF. George Michaelson 13:44 So there's a specific draft in the working group that you have done a project on, and you said that this is called NOTIF, HTTPS. Bharadwaja Meherrushi Chittapragada 13:53 Well, that's the short name. The long name would be like HTTPS transport based for YANG notifications. HTTPS based transport. George Michaelson 14:02 So Vartika, can you just explain a little bit about what the NETCONF Working Group looks at? Vartika T Rao 14:08 The NETCONF working group, or network configuration working group, as long name, works basically on collecting network telemetry data. And so NETCONF is an application layer. Protocol stack, and it works in any transfer base, and it mostly works on changing network configurations. [George: Right] The whole stack is used for mostly changing network configuration. George Michaelson 14:34 So this is kind of funny that mostly when people talk about protocols, you think in terms of bits on a wire, bits in fiber, radio signals, network layers, transport layer. You guys, you're right up the top of the stack here. So Hayyan, this idea, you say it's a YANG protocol. YANG is a protocol, right? Hayyan Arshad 14:54 It's basically a data model. [George: it's a model] It's useful to for network configurations. George Michaelson 15:00 And this is like a textual description of things and... Hayyan Arshad 15:03 not really. It's it's a data model. It can be encoded in any way, George Michaelson 15:08 right? So if I were looking in a draft, I would see a layout of a structured statement about the configuration. And then there are tools that compile this down into a binary form or a structured form. Hayyan Arshad 15:21 So for validation, there are a lot of libraries which we use, so it depends on your the programming language you're using. For example, if you're using C, there's a libYANG library to do it, and for Python, there are multiple like PYANG and YANGSON. We chose YANGSON for our project, George Michaelson 15:40 so YANGSON ... Hayyan Arshad 15:41 I feel it combines YANG and JSON George Michaelson 15:44 So Sid, can you talk a little bit about how you guys were approaching this problem? What is the structure of the system you were developing? Siddharth Bhat 15:53 I mean, the draft defines a publisher, a receiver, and the communication in between them to understand what is supported, what is not supported, and then send notifications. So the first challenge, I think, was understanding YANG data model itself, and understanding how the draft works and how this fits into the bigger picture of other drafts in netconf. George Michaelson 16:14 So you did some pre work as a group reading about this and getting up to speed with the design ideas you're working on here, and you say that it's got both a producer side and a consumer side, [Sid: exactly]. So Vatika, how did you code this? What was the approach you took to coding the solution here? Vartika T Rao 16:32 So our final implementation as the collector, we have HTTPS flask server, and on the producer side we have a simple Python script. George Michaelson 16:42 So flask is a development environment, which is in Python, my favorite programming language. As I say every PING episode I can you previously had some code in C, I think you said Hayyan, Hayyan Arshad 16:56 so when we were working in IETF 121 Dublin, we participate in the hack event there as well. Meher was there on site, and we three were working as remotely, right? So that time, initially, when we got the project, we're using Alex Wang's code for the same [George: right] HTTPs is more difficult, George Michaelson 17:16 but you then came to this one and decided, let's read using Python? [Hayyan Yes]. So Meher the Python system, the flask system, that was the consumer side, or the producer side, Bharadwaja Meherrushi Chittapragada 17:29 That's the consumer George Michaelson 17:30 right? And the producer side, you've got masses of data coming out of these systems, right? How were you handling the data? What was the fundamental model you use there? Bharadwaja Meherrushi Chittapragada 17:40 Right? So the data had to sort of pertain to a YANG model, right? And in the draft, there is no specification as to which model to pertain to but what we use was interface data, which is from another RFC, 8343, which collects information about the interface, the number of packets going through them, George Michaelson 17:58 right? If it's so basic telemetry, and then you would see in network configuration, and you're using this YANG data model to define how to put it into an information system that's but you need them to actually marshal the data. So what did you use as a mechanism to hold that data? Bharadwaja Meherrushi Chittapragada 18:15 Okay, so what we've done is we've written a Python script which reads the interface data from whenever it happens, and that is sent to the collector. So for having multiple producers, we spun up multiple threads on the same system in sort of a network namespace, different network namespaces, so that, yeah, the processes don't, you know, collider stuff doesn't happen. So we have two separate network stacks. One, we run the producer. One, we run the collector. That's the test bit. George Michaelson 18:39 Vartika, how did you implement the message store behind this data flow? Vartika T Rao 18:46 on the collective side, to store the actual notifications that you get from the publisher, we've used Kafka and a TSDB. So to store the notifications, we've used the TSDB called influxdB, and before the data actually flows into the influxdB, we have a Kafka Message Broker in between, George Michaelson 19:05 right? So this is a classic publish, subscribe, information processing model that's a really quite high level building block to use to do things. I think it speaks to how far the technology has come that you can just take a service like this off the shelf you didn't look at you. Using Redis or anything you were happy with influxDB is that like a time series DB, Vartika T Rao 19:26 it's a time series database, and we've taken inspiration of this whole for this whole architecture, from something called YANG push, which netconf, as the whole working group is working towards. So YANG push is basically architecture to store network telemetry data and as an easy way to analyze it later on. Hayyan Arshad 19:46 So I feel Thomas Graff was Thomas Graff and Benoit so we went through a blog written by Benoit Claise [George: yeah] and it mentioned this young push architecture, and I feel they wanted to move towards data analysis, big data analysis through by storing this data. George Michaelson 20:10 Yeah. So you actually had quite good interaction with members of the netconf working group. This was a nice collaboration. [Hayyan Yes] So Sid, can you talk a little bit about what the outcomes are? Have you done a final presentation back to the working group, Siddharth Bhat 20:25 yeah, so we did successfully present what we've done over the couple of months, plus what we did the hackathon and the net count Working Group at ITF, 122, the results are that we did implement what was mentioned. So we have a collector, we have a publisher, and as per suggestions from Kent and HTTPS draft authors. We also did some analysis of the performance you get on the different kinds of encoding, George Michaelson 20:48 so looking at the amount of buffering and [ Sid: exactly] delay between collecting the data and publishing the data [Sid: exactly]. Oh, it's really very good. So Maher, do you feel there might be a follow on project? I mean, you're all fourth year, so you're going to be moving on into other fields, but might there be more work for NITK students here? Bharadwaja Meherrushi Chittapragada 21:07 Yeah, I think definitely we ourselves have our thoughts in mind of what we could do next. Currently, what we've done is we've just used JSON, XML and CBR encoding. CBOR wasn't present in the draft, so we propose the draft ourselves. George Michaelson 21:20 So you have proposed the draft to use a CBOR encoding to do this data. CBOR is fascinating. We talk about it a lot on PING when we're talking DNS, because there's a DNS collector Model C DNS, which is a refinement of CBOR. So you're going from YANG representation to structured data collection. You've currently got it going into JSON, and you're now looking at CBOR, which would be a much more compact encoding, wouldn't it? Bharadwaja Meherrushi Chittapragada 21:48 So from the results of the experiments we've done, we've seen that on network lower bandwidth networks, CBOR performs gives really high throughput. So that's exactly why we're and it's also widely standard, like you mentioned. It's used elsewhere as well. George Michaelson 22:00 So that's future work that someone else might work on. Or are you guys going to pursue a draft maybe? Bharadwaja Meherrushi Chittapragada 22:05 we are going to pursue a draft on that. But there is more work. We thought we could also compress the JSON exam without using CBOR. George Michaelson 22:13 Yeah, you tried, like, libz compression, [Meher yeah]. And that was, that was okay. What was that like? Bharadwaja Meherrushi Chittapragada 22:18 that is the future work we're thinking of doing, right? George Michaelson 22:23 So just to wrap up, do you guys already have an idea where your career path is heading? Do you have have you been hired already, or are you looking at doing studentships masters? What's your plans? Vartika T Rao 22:34 I think all of us have been hired, and at least my plan is to work for a couple of years, maybe find out something I'm very interested in, and maybe pursue a master's on that, George Michaelson 22:44 but with your input statement that you've got a component of ML in what you're doing, it's possible that's more interesting for you. Vartika T Rao 22:51 Yeah, I have minors in machine learning as well, so I am looking at machine learning as one of the things I would be interested in. But right now, networking is on top of my mind. George Michaelson 23:02 So Hayyan, how about you? Hayyan Arshad 23:04 Yeah, same as vertical, I have been hired and will be working for a couple of years and then explore things in the meanwhile. So based on that, I might go in, get into a masters George Michaelson 23:16 Sid, Siddharth Bhat 23:17 yeah, more or less the same, I've been hired, and I would definitely work for a couple of years and then look into like a specific field to do a master's degree in. But, yeah, it's definitely going to be something systems related, definitely not ML. George Michaelson 23:32 I'm an ML skeptic at some levels. I'm not saying the field doesn't exist, but I haven't found a use for it myself yet. So I'm glad to hear there are other people who are thinking of other paths. How about you? Meher Bharadwaja Meherrushi Chittapragada 24:56 Pretty much the same thing, but I will do masters, though. I'm not sure, not, not sure which field yet I am hired, I spent time exploring as well? Yeah. And then figure out what to do, George Michaelson 24:02 guys, I want to say it's been a blast. It's been lovely talking to you. Thank you for coming on and talking to us about this activity. It's great. Bharadwaja Meherrushi Chittapragada 24:02 Thank you so much for having us. George Michaelson 24:05 Finally, let's hear from Professor Mohit Tahliani. So Mohit, thank you for coming and talk with us. I've just done two interviews with two groups of students that you have conducted into the IETF, can you tell us a little bit about what your ideas are here? Mohit Tahiliani 24:25 Sure. Thanks, George, first of all, for having discussions with my students. So I brought these students with me to IETF, because as a part of the course that I teach in my university, George Michaelson 24:35 that's NITK Karnataka, Mohit Tahiliani 24:38 yes, the National Institute of Technology Karnataka at Surat Khan, George Michaelson 24:42 High tech hub of Mangalore, close by. Mohit Tahiliani 24:46 yes, so we are very close to the Arabian Sea. Our campus is right across the beach. So I teach computer networking courses in my university. I start interacting with them when they are in the third year of engineering, and they work with me on multiple courses of networking. George Michaelson 25:01 So you've brought students into a quite strong, functionally loaded, difficult meeting cycle here at IETF, a place where people are often extremely blunt and direct to each other. It's really throwing them in at the deep end in some ways. Mohit Mohit Tahiliani 25:17 Yes, the reason I still thought I should get the students is primarily because my students are more interested in implementing the internet drafts, rather than immediately start working on writing the draft. George Michaelson 25:31 And I might add, they're more than up to this job. I mean, these, the group I just spoke with, have done a complete implementation from a YANG spec, [Mohit: yes] to a functioning system using all the tools in the armory, you know, Python codes, flask, Pub Sub models, everything. It's fantastic. The other group, they're actually working in the network stack, implementing queuing disciplines in in protocols, comparing them, analyzing them, presenting to the group. I think it's fantastic. These guys are doing really well, but there must be a sort of higher goal here. What's your idea about the potential outcome from this? Mohit Tahiliani 26:08 One thing on a very high level, one of the goals is to also give feedback about these implementations to the people who are implementing internet drafts or writing internet drafts, early feedback to the authors of the internet drafts, as long as students are concerned, I want them to go back to the university and tell other students about how was their experience of IETF, and also motivate the younger generation to be at least be aware, George Michaelson 26:33 right? so that you've become part of the supply chain for future brains in a framework like IETF, this is building the pool of new, younger minds working in the standards process. This isn't the first time NITK students have been involved in IETF, one of the students earlier today said that they were in Dublin, but I believe Nalini Elkins has also led a group here working in IPV6 measurement at previous meetings. Mohit Tahiliani 27:00 That's true, and those are my students who have been coming with Nalini to the IETF, George Michaelson 27:04 So it's part of a wider collaboration in some ways, Mohit Tahiliani 27:08 yes. So we have a project which is funded by APNIC foundation, which is related to migrating our campus network to IPV6. And Nalini is one of the IPV6 experts at IETF, and she has also written a couple of drafts related to IPV6. So she was involved with us as a consultant and helping us design the initial phases of how do we migrate our campus network to IPV6. So she, as well as Michael Ackerman, both of them have been working with us and helping us to migrate our campus. So I used to send students with them because I wasn't able to travel to IETF for some different reasons. So they would take care of my students and put them in the right working group, or help them present in places like IEPG. The students have made presentations in V6OPS. The students have presented in Brisbane during the EBPF Working Group as to how we implemented the IPV6 extension headers in EBPF. George Michaelson 27:59 And this time around, they were in netconf and they were in CCWG as well. Mohit Tahiliani 28:04 They were in NMPV they were in TSVWG, George Michaelson 28:08 yeah. And, of course, the hackathon, so they've been participating, integrating with core IETF working group participants. [Mohit: Yes], I think it's truly amazing. MOhit, thank you. It's fantastic. Mohit Tahiliani 28:19 Thank you. I want to share one more point that when we were at the hackathon this week, a lot of people came to our table where we were doing the hackathon, and they were pretty appreciative about the effort that the students were putting and in fact, during that time, they offered two slots to the students to come and talk in their working group. So one of the chairs from CCWG came and said, Would you please present this, what you've done in one of our meetings. Yeah, it was very kind of them to give us the opportunity. And students were really thrilled. And they meet the most out of it, we also got NMOP chairs, telling us that if you could present this, it will be great. So students, again, took that opportunity, and I'm very grateful to them to give us these chances. George Michaelson 28:59 So I feel it's a little like fishing in the pool that you have to expect you put quite a lot of energy into a group like this, maybe only one or two will decide that this is a future, long term outcome for them. But you know what? I think it's worth it, even if it's at a low percentage outcome in the longer tail. This is an amazing way to see people come into what we do I really, seriously want to thank you, Mohit I think this is a great initiative. Thank you. Mohit Tahiliani 29:25 Thank you. Thank you. That validates all the efforts we are putting and we are very happy to have support from all of you. In fact, the two students with whom you have earlier done our podcast, both of them are still in the field of networking, George Michaelson 29:36 great, Vanessa. And Kavya, yes, Mohit Tahiliani 29:39 and Vanessa is now working as a network engineer at meta in London. [George: Oh, wow]. And Kavya is pursuing her Master's also in the field of network, right? So I believe that someday we will have these two students come back to IETF because of their long term association in the field of networking. And I agree it's just bringing students and hoping they would stay back. George Michaelson 30:01 That's great. Thank you. Mohit Tahiliani 30:02 Thank you George for doing this. George Michaelson 30:06 If you've got a story or research to share here on ping, why not get in contact by email to ping@apnic.net or via the APNIC social media channels. Also remember the measurement@apnic.net mailing list on orbit is there to discuss and share relevant collaborative opportunities, grants and funding opportunities, jobs and graduate placings, or to seek feedback from the community on your own measurement projects, be sure to check out the APNIC website For all your resource and community needs until next time you.