Cloudflare TV

People Behind the Packets

Presented by Nitin Rao, Jayshree Ullal
Originally aired on 

Tune in to hear conversations with leaders building the infrastructure the Internet relies on.

This week's guest: Jayshree Ullal, president and CEO of Arista Networks.


Transcript (Beta)

Hello and welcome everyone to another episode of People Behind the Packets. My name is Nitin Rao and on the show we speak with industry leaders we respect who are changing the Internet as we know it.

I'm honored today to have as my guest Jayshree Ullal, President and CEO of Arista Networks.

Jayshree has really seen the evolution of the networking industry and taken Arista really from zero to a multi -billion dollar business and so it's a real honor to get the opportunity to speak with her.

Cloudflare is a happy Arista customer. Thank you so much for joining us Jayshree, really appreciate it.

Thank you Nitin, it's an honor to be here. Cloudflare is one of my favorite friends, partners and customers and congratulations to you too.

You have had a phenomenal success and it's proud to watch you guys start at the bottom and grow the way you have.

Thank you so much and we appreciate the partnership.

So there's so much to speak about and I've promised to not ask questions about COVID because we've all spoken so much.

COVID is that five-letter word you can't avoid.

Yeah, you can't avoid. Like Trump but oh well. So Arista now has equipment really deployed all over the world.

It's an iconic networking gear brand.

Well before that you were at Cisco for many years and before that in various engineering roles.

I was wondering if we could actually start if you wouldn't mind from just the earliest years as an engineer.

Would you mind talking about folks who inspired you and what you learned and the experiences that led you to the role you are in today?

Yeah, well we won't go back to where I was born in London and then you know I grew up in India but my early years in the involved in companies like Fairchild that don't even exist or got bought building silicon, high speed silicon for mainframes and memories and then I went to AMD which is of course now a very successful company which was my first time that I built networking chips chipsets.

Back in those days the highest speed was 100 megabits FBDI and you know and back in those days the process geometries which are today's seven nanometer was you know 100 micrometers so these were 10,000 gates not million transistors like you have today but nevertheless the one thing I want to do want to tell you is I got to see Silicon Valley when it was really silicon and when it was really fruit orchards so that was a real honor and some of my early influencers were the semiconductor guys who actually built up this Bay Area and you know the and these folks you know they were not they were not known for humility but they were known for intelligence and gravitas for sure and so if you look at AMD that was founded by Jerry Sanders he taught me a thing or two about not how to be a good engineer because that was not his background but how to take it to market and influence customers which I think was very important because as you all know we're geeks at heart but you're often selling to very mission critical CIOs and CISOs who you have to be able to explain and simplify this to so the only thing I didn't like about the semiconductor industry were many things that it's hardware at its core and you hear all about software these days but software has to run on something and the product cycles for that could take anywhere from you know three years to five so by the time you conceived a product and you actually saw it to market the market could change so that's when I went into the systems industry and networking and my career started with a company called Langevin Bass where I built some of the largest switches and routers in fact I installed the FTDI backbone at Microsoft Redmond and Cisco was a five million dollar company then and Langevin Bass was you know 10 to 50x bigger and I still remember when we were actually reselling Cisco's routing products because Langevin Bass did XNS routing and Cisco did IP routing but the rest is history Cisco went on to become much bigger.

I think that the thing to take away from all this is for most part networking until the last few years has remained a connectivity mission but that has really changed with the advent of the cloud and Cloudflare and Arista.

I think the migration from just basic Internet and worldwide web connectivity to today you know not just being the Internet but being the cloud provider has really you know been the most rapid change I've ever seen.

I didn't think it would happen this fast in my lifetime but I think we saw 100 years of work happen in 10.

It's incredible to see that evolution from you know 100 megs to like now we're talking about you know 400 gig equipment and it's crazy and more.

What was special about you know Fairchild and SGI that like alumni have gone on to start off so many interesting companies?

They have. If you look at one of the best examples of alumni is actually Jensen in NVIDIA.

He was with me at AMD and I think what semiconductor industry builds for you is a certain foundation of technology, frugality.

We had to do more with less you know every dollar mattered, every transistor, every cogs mattered right and then I think there was an element of just work ethic too where again this was an industry that came out of nothing and we never knew what the next step or next navigation point would be.

So living with uncertainty, being deeply technical and then being extremely cost-conscious are some of the things that the semiconductor industry really taught me.

Which of the companies that, so of course Arista, which are the companies that still are in the Bay Area but keep in touch with hardware?

It feels like it does feel like everyone just starts off with an AWS instance and doesn't really talk about hardware.

Yeah it's certainly not fashionable to say I'm an HDN, hardware defined networking company.

Everybody says SDN but you'd be surprised that even if you look under the cover of all these companies, even the big ones like Cisco and HPE and Dell, there's a substantial amount of hardware built right.

It's the monetization and the investment is much greater in software, no question about that.

Because more and more of us back in the day when I was in Cisco, we built our own ASICs and chips.

Today that migration to merchant silicon whether it's with Intel or Broadcom or WAGO or Marvell etc has made us all less dependent on designing the latest FPGA or chip.

But what we still have to do is focus on how do we twiddle the bits on that hardware, whether it's on the compute plane or the management plane or the data plane to make our software come alive.

So I think you know put it differently, software is definitely the name of the game in terms of investment and engineering but hardware is a very critical gift wrap.

And through this conversation I'd love to also talk about sort of Arista's approach to software that I think has really distinguished it from other companies.

You know so true, when I joined hands with Andy Bechtolstein who as you know is a legend and a luminary who founded Sun who invested in Google and he had funded the company in a very unusual way in the first five years where the company took no venture capital money.

So it was pretty much Andy and David who put in their money but they did something very unique which is not only were they funders but they were also founders.

So you care deeply not just with your pockets but you care with your brain as well.

So at first when Andy approached me and he and I worked together I bought his company in Cisco and he worked with me for seven years when Granite came on.

So he said I'm looking to do this networking company.

My first reaction was oh not another networking company because you have to remember this is way back in 2008 when it was game over.

Cisco was buying up every company and the future was all social networking like Facebook or Google or search or you know.

And I think you were a senior executive at Cisco.

I was although I have to tell you I left my Cisco position before deciding what to do because I think it's very difficult to stay in the job and do a job search.

So I actually thought I'd get away from networking and go pursue clean tech and I looked at battery and solar and a lot of technologies and ultimately realized when you're as old as I am you have to rely on my expertise and what I know.

I can't become a PhD in you know clean tech and so I was spending a lot of time learning it but I didn't have the wisdom on it like I do networking.

And what Andy pointed out is look we can put a lot of energy on the silicon but we can rely on merchant silicon and I'm going to put our energy into the software stack.

So what drew me to Rista was the people and the technology.

So if you look at the Unix then database world they had all moved to publish subscribe models for software.

The networking world was so busy just running fast and building features.

It was all a monolithic blob of spaghetti code where everything was inter-process communication.

If one agent failed everything came down with it like a Christmas tree light.

So this publish subscribe model has turned out to be one of the biggest advantages of Arista's architecture that at that time was a million lines of code.

Today it's 50 million lines and can somebody else come and build it?

Yes of course. Can somebody perfect it? It would take them 10 years. So I think the industry missed the boat in networking and they kept thinking okay that's for the enterprises they don't need performance and good enough is good enough.

But best of breed became very important and two customers really drove us there.

What we call the cloud titans and these are building some of the largest scale networks.

You know they're putting in zettabytes of storage 100,000 servers is nothing for them.

A billion work you know virtual machines and containers. So companies like Amazon, Microsoft, Google, Facebook and even second tiers like Apple and Oracle and SAP etc recognized that in order for them to you know be competitive they had to not only they had to be a cloud and then sell the cloud.

So quick story for you on that.

Microsoft is now well known to be our largest customer. They seem to be my good luck charm.

I told Satya that and you know in that I installed the FTDI backbone and I won them again in Crescendo and then in Cisco and we're fortunate to have a partnership here.

But two weeks before the company was launched somebody sent us a legal notice and said hey our original name was Arastra Networks because we were based on Rastradera Road.

They said hey we don't like your name it's too close to our name which is Astra Networks.

So we were a young company I was only 30 engineers then so we decided we'll change the name and our first choice was Azure.

So we called up the domain name for Azure and it turned out somebody else and you know who the somebody else is had taken it two weeks ago before us.

So and then we ended up with Arista. Arista for those of you who don't know is also a great records company for music distribution.

So even today we get a lot of auditions on our Arista email asking to you know try out their music so I could start a side business on music distribution too.

But Arista and Azure became very intertwined.

That's just such a wonderful story. Do you think as a word so one of our guests a couple weeks ago was Pradeep Sindhu and we were talking about the evolution.

Do you think like as a word is networking gear sort of appreciated enough?

Like do we do folks understand what happens behind the hood or should they?

And you know it's like do you need to understand if you drive a car how the Tesla's built and how the or how your hybrid or even a regular old car is?

Not unless you're a car mechanic and not unless you're really interested.

So I think it's very similar.

Is networking appreciated? Not until it goes down. Right?

That's true. But you know so you're a hero as long as you keep it up but actually you're a silent hero and you're a zero very quickly if it comes down.

Right? So when you look at the data center I think there's always the you know four three two one problem which is four dollars gets spent on compute, three dollars on storage, two dollars on technologies like yours and virtualization you know software let's just say and only one dollar on the network.

But that one dollar better be mission critical.

And it holds up everything else. Sorry? It holds up everything else.

Absolutely. So to your point about does the network get enough appreciation?

No. But does it get attention if it goes down? You bet. Right? So people have realized particularly in the cloud how mission critical it is and even in the enterprise they don't have resources to manage it and operate it but they know that they need to be responsible.

There used to be a time you you didn't get fired for buying IBM or Cisco.

Today you will get challenged if you don't look at alternative technologies especially cloud-based ones.

Right? So networking couldn't be more important but it's very much behind the scenes and I think companies like Arista and Cloudflare have made networking sexy again and they are really addressing the root problems right now which is availability, agility, automation across the network, change control, compliance, security, analytics, telemetry.

So networking by itself not getting appreciated but igniting networking to do all these functions you bet it's very important.

I thought it was interesting that you mentioned compliance and as an industry it does seem like there is the set of companies that are sort of pulling back but there's a set of companies that are actually leaning into it and and identifying the opportunities to help customers.

Yeah even we are we didn't think of ourselves as a security company but we just recently bought a very advanced threat hunting technology that was very adjacent to what we're doing and what you're doing which is the the perimeter between you know the boundaries or where you put a firewall and what workloads go to the cloud and shadow it and the introduction of iot and how one user could equal equal five devices.

I got my iPhone, my iPad, my iWatch, my maybe my jewelry which will have an IP address.

It's just changing the way remarkably the way we live and the attacks are more and more malicious right.

So here's an interesting statistic 50% of the devices we all run around with are unmanaged and another 50% the malware and maliciousness is never attacked and never detected.

So I think we've really got a compliance problem and as you know it's a board level topic but in my view we're playing whack-a -mole there instead of really getting the to the root of a good foundation that provides you that predictive proactive autonomous threat hunting which I know it's no talk is complete without saying AI and ML but it really is a combination of using a predictive AI ML nucleus but also expert systems, human expert systems.

Do you and I know that you can for example you know now no provision you know Arista equipment in the cloud so you've got sort of work you know virtual equipment.

That's right. Does the relationship from the perspective of a company like Arista like will networking companies have access to more customer data over time than the traditional model of?

That's a really good question.

I think the networking technologies are becoming more and more data-driven right and I can look deep in the package just like you can at layer two, layer three, layer four, you can go even deeper as you know.

Now what we do with the data and I think our approach to this unlike the social media or social networking companies would not be to eyeball the data or carry the users and the identities who would value their privacy but what we can do is provide more data-driven networking where you can get the anonymized data sets and do some really cool pattern matching and put this through a time series database like we have on our EOS software and our cloud vision and really give them some predictive analysis.

So using it to give the CIO or CISO the right trends is what we think is fundamental particularly like you say as workloads are moving to the cloud, they're getting containerized, they're in the premise, they're hybrid and now with you know with the work from home and transit and user it's even more difficult.

It was interesting one of the things we are doing even in our Wi-Fi gear is we're giving a quality of experience where we can prioritize whether the Zoom traffic is more important to you for example than the Webex traffic.

I just say that, happen to say that and in fact one customer asked me does that mean you drop the Webex traffic and I go no no we can prioritize it but that quality of experience is more important so I think not getting the data but acting on the data is fundamental.

We're switching gears a little bit for folks who are considering becoming network engineers where you know if I'm in college and I'm just graduating I'm thinking about what to do, why should I become a network engineer?

Because that's where you get the deep combination of technology, engineering and you know I think real fundamentals.

Back when I was going to school it was all electronics and a few computer science classes and I had to learn networking in my master's because it wasn't there in my undergrad and you know and I guess I'll start with a story there.

We had an engineer, a PhD who was doing some fantastic work on BGP routing at Arista and then two years into the job he said he got a really good offer from Google and I said well you know do you like the work here and you know it was Arista was pre-IPO so he would have done financially even better here than Google and he said I really like the work here but you know this company is doing some cool video search and you know that's a more front-end use case I really want to go do that but it wasn't a case of he liked one job better than the other but one appeared a little more cool right.

So another such engineer also went to Yelp but both of them came back to me later on and said the work they did in Arista was much more fundamental, they learned much more, they went through the complexity of our algorithms, our debug, our verification, our automation, our innovations and you know had they not had that you know they wouldn't have been a good engineer.

So I think there's a you know think of network engineering as almost a fundamental that'll help you in any company right because it's where hardware, software, algorithms, dev tests and higher levels of applications meet up.

If you go straight to one of those higher levels it's not clear you'll ever be a fundamental.

Now I think it's also important for companies like yourself and mine to offer the right internships and offer the right practical training because there's only so much you get out of textbook and that's something Arista's investing a lot on probably just a little less this year with the work from home but a majority of our investments go into internships and new college graduates because I think that's where it starts when I look at how I got influenced right and that's where it began.

By the way just to end that Google story, ultimately he made the decision to move to Google because he was a single guy and I asked him what does your mother want and he said my mother wants me to go to Google because I will get fed every day well.

So I realized the decision point is food as much as it is anything else right.

Sure. You're on the board of Snowflake and they've done terrific and I was reading an article about the intensity in organizations.

Is there something special about companies like Arista and Snowflake where you can just get a lot done in a short period of time?

Yeah that's a really good parallel. You know Frank Slipman who was my peer in ServiceNow, I've known him from his data domain days and ServiceNow and then he's just you know I'm so proud of what he's done at Snowflake and not just what he's done but the founding CEO Mike Spicer took a very unusual path just like Arista did with how to fund a company.

He wasn't just a VC, he wasn't just a funder but he was the acting CEO and then when he came from Juniper and Microsoft but to come back to your story I see some parallels.

I see a certain driven energy in both companies where they're not just managing, they're leading, they're saying I want to go there, I want to change the face of data warehousing and databases and there's a level of conviction sometimes without data sometimes with data where the whole management team and leadership team has to believe, has to be united and I think you'll eventually get the data to back it up but the conviction is as important as the data.

The second thing I see is a level of leadership where Frank's team and Arista have shared that in common too.

You know we don't manage, we lead.

So our engineers and our sales leaders and IC leaders, I ask them all to think of themselves as mini CEOs of their sphere.

Don't ask your boss what you do, think what your boss would want you to do and 90 percent of the time you probably already have the answer and of course children need parents and schools need principals and countries need presidents but on a day -to-day basis if you can be empowered to lead and do your sphere or sandbox well then the collection of those sandboxes becomes a great company.

So I see that and the third thing I see you know people always say to me you know is Arista market driven, customer driven, engineering driven.

There's no doubt that we are built by engineers for engineers but I think at the end of the day you need the confluence of all of that.

You sometimes need to be customer driven, sometimes you need to be able to tell the customer nope that technology won't work, you're telling me to go a wrong way even though you're always right and you also need to be very quality driven.

I'd rather apologize any day and every day for not delivering a feature but I never want to apologize for bringing down a network right.

So I think some of these fundamental principles are not in all companies but when you see that common value system, culture, disruptive technology and leadership skills then I think you know you have a home run and then obviously it helps to have a market and momentum and all of that timing and luck but that's just icing on the cake.

The fundamentals have to be correct.

Yeah and congratulations on building such a terrific company.

Can you talk a little bit about how your leadership style has evolved over time?

So how would you contrast it to 18 years ago? I think it's a very good question.

I think most people would say my leadership style is a combination of sometimes command and control and then sometimes delegation and obviously when you're a young company you have to keep every minute detail in your head and at the same time I've always believed that leadership means hiring people who are A players right and then if you hire the A players then every once in a while you may have to say to them hey I need to steer you in this direction but by definition they will do a work.

So probably the biggest you know piece of my leadership is empowerment.

Also wear the passion on your sleeve. Don't wait to extrapolate and get every point of data.

Make decisions. I encourage the leadership team more and more right now to make decisions and then course correct rather than get paralyzed in making a decision right because now the course correction is important because nine out of ten may be right but you've got to make sure that one you're able to say I'm wrong.

I didn't do it correctly and it was a fork in the road and I went the wrong way and I want to come back and I find myself doing that even more.

I find myself a little frustrated with some aspects of my leadership today where I just don't know enough about what's going on and I'm the kind of person who's wired to but I've had to adapt and learn that very very well.

So there are areas I spend more time on.

Strategy. Where does the company head the next three to five years?

What's the right M&A? What are the right leaders we need to hire for that three to five year journey?

A third of my time is always in front of customers.

That's a key part but I used to spend a lot more time on technology than I do now.

It's probably only 20 percent or less of my time now but that's again a natural process.

At the same time I don't think as leaders we need to turn into professional managers.

I think we should lead with our expertise. So I'll still jump in, write my own blog or write a white paper or contribute in a certain way mostly for my own therapy.

The mind is an important thing to to keep going. So we have just a little bit of time left.

Maybe looking forward what are you most excited about the next 10 years?

I think that buzzwords aside the United States and Silicon Valley has already demonstrated that tech is a very important foundation of our future and not only is tech influencing technology but it's influencing the way we work live learn and play and there can't be a more better example than how we're all using the cloud.

But I think the last 10 years I would say was about how fast the cloud came upon us.

The next 10 years is going to be more of a balance on what makes sense to use in the cloud and maybe we'll go full circle again and say those cloud native apps were great but some of those cloud principles need to be brought back into our daily lives at home or in the enterprise.

So I actually see it as a six of one and half a dozen of the other kind of problem.

I think the return to a more balanced workload between cloud and premise is going to come upon us and that doesn't mean the cloud vendors will go away or that they won't participate in that but you know we when I talk to a lot of customers they still have their mainframes.

So the ability to go all the way from client to cloud including the mainframes that they may have had from the 80s and 90s is very very important.

I also know everybody loves to say software eats our hardware.

I'm a big fan of building the right software but a software has to run on something.

So I would hate for the semiconductor industry to get gobbled up and consolidated like it is.

I think the innovation and pushing the geometries you know and getting intel back to being a great company again and all these semiconductor companies that made the silicon valley who it was in the 80s and 90s need to continue to push the envelope.

Great, well I really appreciate you're joining us and your candor.

Thank you so much this has been a real honor. Thank you for having me and thank you for prompting all these great questions.

Have a great day.

You too. And congratulations to Cloudflare. Thank you. you