Cisco Optics Podcast Ep 59. Why some optics are good for AI and some aren’t (1/5)

[00:00:08] Hello everyone and welcome back to the Cisco Optics Podcast where we talk about pluggable optics for networks. I'm sure it's no surprise to you that AI has been steadily changing the world, but did you know that optics is a key part of its hardware infrastructure? To explain it, fortunately we have a seasoned product manager who knows both the switching side and the optics side. And lucky for us, he sits right next to me at the office and agreed to chat about it.

[00:00:34] In episode 59, we begin a new conversation with Paymon Mogherabi, Senior Product Manager at Cisco's Optics team, also known as the Transceiver Modules Group. We begin with his start in Cisco and the history of data center hardware architectures. Paymon Mogherabi is a network industry and Cisco veteran of nearly three decades with electrical engineering degrees from UC Irvine and USC. After starting at Cisco as a Technical Assistance Center engineer, he became a technical marketing engineer for Cisco's Catalyst switches.

[00:01:03] He then took product management positions for Cisco's Edge Services router, Nexus data center switches, and UCS server products. He is now a Senior Product Manager in Cisco's Transceiver Modules Group and has sat next to me for the past seven years focusing on data center applications. And now join me as I talk with Paymon Mogherabi.

[00:01:45] Hey Paymon. Hi Pat. How are you doing today? Good. How about yourself? Good. Good. So, um, yeah, why don't we start? I, I, I, I feel like I should know some of this already, but, um, what, what got you into Cisco? How'd you get into optics? Can you just kind of run through that story real quick? Sure.

[00:02:09] Um, so, uh, for most of my career within Cisco and also outside of Cisco, I've been primarily, uh, on the switching and routing side, um, whether it's been campus or, or the SP environment, but, but mainly on the data center side. So, a lot of, um, work on the data center side, both as a product manager, as well as a technical marketing engineer. Um, and, um, actually I didn't know you were a TME. I was a TME. I was a TME. I was a TME for about four years. You, you were in TAC for a while, right?

[00:02:38] I was in TAC as well. Okay. I thought you went from TAC. I guess I didn't know. That was my first stint. That was my first stint. Okay. And so the first time, uh, at Cisco, yeah, I was in, uh, I came in, uh, to TAC. So I, I firsthand understood all those issues that customers face and all those, uh, And Cisco TAC for people who don't know is? Technical assistance center. So that's basically who you call when you have, uh, uh, Um, issue. Difficulty with something, right? Yes.

[00:03:04] When you have a problem, uh, network connectivity, you call into Cisco TAC. Yeah. And then I spent a few years as a technical marketing engineer on some of Cisco's older, uh, some of the catalyst, uh, switches, uh, the cat 6k product line. Um, then moved into product management role. Um, again, primarily on switching, um, some campus, but then I moved into the, the nexus world where it was, it was all, uh, data center.

[00:03:34] So you were both TME and a product manager for catalyst. Yes. Correct. And, um, on, uh, on the data center side, uh, with the nexus product line, um, a lot of exposure to data center, of course, customer customers, um, learned a lot about, uh, the, the, the difference in, in a data center model versus maybe a campus world. Uh, in terms of speeds and feeds in terms of, uh, product refreshes.

[00:04:00] Um, I then made a kind of a jump towards, uh, compute, went over to, uh, UCS product line at Cisco. Uh, and, uh, whole different world, whole different world, uh, whole different world. Uh, uh, all in compute. Um, but these, these, these UCS products still went into data centers though, right?

[00:04:48] Okay. And, uh, uh, the thing yet. Cause I thought we, even data centers still have the three tier, the three layer. Correct. So it was going through a transition. So prior to that, it was all about effects architecture. Oh yeah. The, the nexus 5k, 2k, and essentially the 2k being like an extended line card coming out of a nexus, uh, 5k.

[00:05:09] Uh, but that architecture started to slowly go away because customers were starting to move towards a fixed architecture, a fixed switching architecture. Fixed meaning? Um, fixed switches. So 32 ports, uh, so not, not line cards with a chassis, but like a box. Uh, more or less, uh, so I guess you can say both, um, with the, with the modular switches, uh, getting pushed to the, to the, maybe to the spine or in an end of row architecture.

[00:05:40] Sorry. Just to the definition of fixed versus modular switch is what? Uh, so with the fixed switch, your, your, um, of course your ports and your route processor and the fabric are all within a, within a switch within your, within the box and they're not fruable. So you cannot pull out a, a line card or a port. Uh, you cannot pull out the switch fabric or the, or the, um, route processor. It's all within a pizza box.

[00:06:07] It's all one box and there's nothing, there's no modules to pull in and out. Correct. Okay. So typically the, um, the refresh cycle of the fixed switches is faster than versus a modular because modular, you're looking at a chassis that you can swap out line cards. Right. Okay. So, um, next generation of line cards come, you keep the chassis. You just swap out the, the line cards. Now there's, of course, it's not as seamless as I, as I make it sound like, but you still, there, it is, you're not stripping out an entire chassis.

[00:06:37] So you still have that, um, you still have to upgrade and replace, but you're not, you're not pulling out an entire chassis. Yeah. So it is, uh, easier, uh, in some sense it's easier if you can just swap out the line cards versus having to take out the entire, uh, chassis. Okay. So the, the model started to change from effects architecture to a more of a leaf spine. Um, and, um, of course the compute side, same, you know, you had the servers, you had the NICs all connecting upstream to the top of rack switch.

[00:07:06] So that was new to me because I, in the past, uh, my focus was entirely on the fabric, uh, on the leaf and spine. But now I had to start looking at the NIC as an entity, as a, as a component that connects the, the compute side of the world to the network. Okay. That's the boundary. It's the NIC. Okay. So, uh, I, I learned a lot while I was with, uh, UCS. Uh, I then made a big change. I jumped into optics.

[00:07:35] Uh, I joined the optics team and, and it was around mid 2017. Uh, and, um, and it was that, that, uh, even though I had some, uh, background as you know, background with optics, but it was more on this optic being qualified on, on this, this chassis. Mm-hmm . I didn't fully understand the whole, um, supplier model, you know, uh, how to deal with, you know, GSM and, and all those, uh, details.

[00:08:02] Um, so got into optics and then, uh, of course, uh, started to learn about the optics, uh, world. Um, and, uh, but more of a focus towards the topology versus getting deep into the optics itself. Yeah, because I mean, it makes sense, right? Because you have all this experience with the various, uh, network architectures. Correct. In both data center and the catalog and the campus networks, right?

[00:08:32] Right. So, yeah, yeah. And so it, it made sense to, for me to focus more on the topology side. Mm-hmm . And I did have more interest in that. Um, but, uh, in the optics world, of course, you're plugging into a topology. So you have to understand the topology. Yeah. To be able to sell, um, your, your product into a, into a, into a use case. You have to understand the topology because topologies are different. Use cases are different.

[00:08:59] A use case in the campus is different than data center or in the, or in the compute side of the data center versus the fabric. So all these things, um, if you, if, um, as you're looking to sell your optics, it's, it, I think it's very beneficial to understand where you actually plug it into. And I, I'm guessing even more so now, right? Because when, I remember when, um, when we first, or when I first met you, you were still in the Nexus team. Correct.

[00:09:27] And yeah, like you said, you were the optics person in the Nexus team, but it was more for purposes of. Qual. Just coordinating what to qualify and all that, right? Correct. Um, at that time. And that was, you said you joined us in 2017. So the seven years ago. Yeah. Yeah. Well, it was a two year gap. I was at UCS for about two years, about two years. Okay. So, um, so, so I met you more than nine years ago, I guess, which seems crazy. I can't, it's wow. Okay.

[00:09:57] Really? Yes. Yes. Nine years. Oh my God. Okay. Okay. Anyway, what I was getting at was back then optics was a much different, uh, how can I say? It took a different amount of thought process. Yes. And, and I would say on top of that, there are just more choice. The decision is much more difficult now because there are so many choices.

[00:10:27] Correct. Today. And back then there weren't that many choices. It was basically, well, how far do you need to go? All right. Give me a number and I'll tell you which optic you need. Right. So in the case of Nexus, we had, you had 40 gig, you had some 100 gig, but mostly 40 gig and 10 gig. And, and, uh, your choice was okay. Am I going 40 to 40 or, um, am I going 40 to four by 10 gig? And that was it.

[00:10:55] But now, as you look at all the different options, you go up all the way up to 800 gig. You look at 400, 200, a hundred. You have all these different combinations. The interoperability, um, options are, are significantly higher now versus what was back then. Uh, back then it was simple one or two. So you want multimode or single mode and, and is it 40 or 10 gig? Whereas now it's a, it's a completely different story.

[00:11:23] So it's much more complicated. So would you say that you're just like, you're always where the action is? Is that? Um, I think from a use case, yes. And I, and that's what I, I mean, I like that is from a use case. Um, we, we are looking at a lot at the more of the, of the AI deployments. Mm-hmm. AI has become, of course, as we all know, uh, is a hot industry. Yep. Okay. So great segue. Wow.

[00:11:58] That was the first part of my new conversation with Payman Mogherabi. Next time we'll get into data center hardware architectures and traffic patterns for AI applications. We have a new website. It's called optics.podcastpage.io. You can either listen there or use the same podcast platform you've been using all along. Please subscribe. Better yet, leave a review, especially if you use Apple podcast. Remember we're part of the Cisco podcast network where you can find other great Cisco podcasts too.

[00:12:27] We also have educational videos on YouTube. Just go to youtube.com and search on Cisco optics. Thank you for listening. This is Pat Chau in technical marketing at Cisco optics. The next episode is part two of my conversation with Payman Mogherabi. Until next time.