8 Comments

Summary:

It is fashionable to obsesses about the web startup phenomenons and forget old fashioned Silicon Valley startups – ones that makes hardware, writes software and along the way clocks in hundreds of millions in sales and profits. And like everything good it takes time to build one.

Andy Bechtolsheim, Arista Networks, at Structure Big Data 2011

A couple of days ago an email from Jayshree Ullal, chief executive officer of Arista Networkspopped up in my inbox. It was mostly to check on my health, but Jayshree also in passing mentioned that at some point she wanted to give me an update on Arista’s affairs. So we talked. And by the time I got done talking to Ullal, I was wondering to myself: Is Arista Networks the next Juniper Networks?

But more importantly, Arista is a shining example of the importance of long-term thinking and patience when it comes to building a big business. In these days of coupons and curation, Arista is a throwback to the Silicon Valley of old, when talented groups of engineers built products that you could see, touch and buy. It’s not based in hip San Francisco or trendy Palo Alto. Instead it is based in the decidedly unfashionable and boring old Santa Clara. The only virality in Arista’s business plan is happy customers telling other customers to buy their stuff. Their customer acquisition strategy doesn’t involve Tweets or Facebook, though as a company they are happy you use Twitter and Facebook and all sorts of cloud applications.

And the hits keep coming…

Andy Bechtolsheim, Arista Networks, at Structure Big Data 2011Arista is the brainchild of Andy Bechtolsheim, Kenneth Duda and David Cheriton. Think of Bechtolsheim as the Quincy Jones of Silicon Valley — he has produced more hits than anyone else. When at Stanford, he helped design a workstation with built-in networking. That led to him co-founding Sun Microsystems.

In 1995 he left Sun to start Granite Systems, a company that was developing high-speed switches, with Cheriton. A year later he sold it to Cisco for $220 million and became the head of Cisco’s gigabit switching business. In 2001 he started Kealia with Cheriton and in 2004 sold the company to Sun, returning to the server maker, who at the time was on the ropes. A year later, Arista was born. Along the way he cut a check for Larry Page and Sergey Brin, which resulted in a couple of zeroes being added to his bank balance. And there are more “angel” investments that have collectively sold for hundreds of millions of dollars.

But that monetary success is beside the point! The point is that when it comes to networks and network infrastructure, Andy is a guy who can peer into the future faster, better and further than anyone else. Cheriton and Duda are no slouches either.

Cheriton is a professor of computer science and electrical engineering at Stanford University and a world-renowned researcher in networking and distributed systems. And when he isn’t busy chalking up patents, Cheriton starts companies, such as Granite Systems, which since has become the underpinning of Cisco gigantic switching business. Duda is the CTO of Arista and vice president of software engineering. Bechtolsheim, Cheriton and Duda worked together at Granite and then at Cisco.

Ullal, who spent more than a decade at Cisco (including a stint as Cisco’s senior vice president of datacenter switching and services), knows the switching business like the back of her hand and is the perfect CEO for this team. She joined the company in 2008.

When it comes to switches, it doesn’t get better than this team.

Big networks, big data needs big switches

Arista has bet on a colossal trend — networked computing.

Jayshree Ullal - President and CEO, Arista Networks - Structure 2011The idea is very simple — as we start to depend on cloud-based applications and services for our computing needs — you know, from trading stocks to business analytics to running cloud apps — we would need a way to move data that has no delays and at pretty rapid clip. And that is not all — these new higher-speed networks would need new kind of operating system software and  networking management capabilities to adapt to the different kinds of the network traffic.

Of course, today it seems pretty obvious but Andy and his co-founders got this idea back in 2004. Thanks to their work at Sun, Cisco and other companies, to them it was pretty clear that there will be more servers, more devices and more networks in more places that would need to be networked. There would be need for a whole new class of switches to make networked computing possible.

The company was started in 2005 as Arastra and has been funded exclusively by Bechtolsheim and Cheriton. This has allowed the company to bypass the impatient Sandhill Road crowd and focus on building for the long term. It meant starting from the ground up, and that included writing an operating system (they call it Extensible Operating System) and using it as the heart of its strategy. Coupled with merchant silicon (chips that you can buy commercially), Arista built a hardware platform that adapts to the needs of the high-speed networks and the traffic. Arista’s approach allows it to boost its hardware capabilities but more importantly allows the flexibility needed when operating the networks.

Today, Arista makes 10 Gigabit Ethernet (10 GB-E) switches. These switches are used to shunt data around on networks at mind-numblingly fast speeds. No, you are not putting these devices in your office or in your homes. These are highly specialized switches that are in demand from financial trading firms, cloud service providers and large webscale companies that need to move data around their various facilities – whether in between cities or inside of data centers at blazing fast speed without any latency and delays.

Just like Juniper

There are many parallels between Arista and Juniper. Both of them are taking on Cisco Systems in its core businesses — routers and switches. Both of them picked a “niche” instead of taking on Cisco head-on.  Both companies wrote their own network operating systems to address the changing nature of networks and do an end-run around Cisco’s aging platforms. Both companies bet on industry defining megatrends.

In 1996, Juniper was formed with the idea that the massive growth in the Internet traffic would need a router and a networking OS that was built for this fast-growing network. Instead of companies buying  routers like they did from Cisco, the demand would come from a new category of router buyer — the Internet Service Provider. And boy they were right. Sure they have their problems today, but fast forward to now and you can see that Juniper is a $10.7 billion in market capitalization company with (estimated 2011) sales in excess of $4.55 billion. When it comes to routers, Cisco and Juniper are now peers.

Arista’s story is somewhat similar. A fast paced and high-energy remake at that.  It is aiming at the high-end of Cisco’s core bread and butter business — switches — and succeeding. The only difference is that Juniper continues to make its own chips while Arista is focused more on using high-speed networking chips cooked up by the likes of Intel and Broadcom. And Arista can be bigger than the router maker — much much bigger — because it is addressing a much bigger and faster growing market.

More networks, more money

Arista has sold its products to a thousand customers in thousand days, says Ullal. The company has grown from 30 employees to 300 in two years. It started selling its switches to the financial community — a majority of the large Wall Street banks and institutions are customers — but since then have expanded to work with cloud service providers and web companies as well. No surprise. According to Infonetics Research, during the second quarter of 2011, the 10G switch business was the fastest growing category, up 19 percent quarter-over-quarter and 58 percent year-over-year.

No surprise, then Arista is on a run rate to clock $150 million in sales this year. And yes it is profitable — enough to not worry about getting another cash infusion from its co-founders. The company steadfastly refuses to talk about the amount invested by the co-founders. They have competition — Dell, Cisco and Brocade are all chasing opportunities in these high-speed switching market — but from the looks of it Arista’s software gives it an edge. For now.

Ullal, when asked about Arista’s IPO plans, pointed out that current state of its business and the “patient” backing of its co-founders means that the company isn’t in any rush. You don’t build a company of lasting value with an exit in mind, she said. Perhaps, that’s a lesson all of us in the startup community should heed to.

  1. Excellent Article !Glad to know that my school buddy is making waves in the clouds!!Manisha

    Share
  2. Silly pet Arista reporter. Juniper rose to 30 percent market share in 2000, and now are down to 28 percent, with a dying set of products and customers. He is not doing arista any favours by comparision. Arista is just another box vendor like dozen others, all feeding off Cisco’s success, and dreaming to be big ‘one of these days’. Long wait.

    Share
    1. I love getting snarky remarks from folks who don’t even want to use their real name and standing behind their arguments.

      Share
  3. Did Juniper take 8 years to ramp to reach a $150M run-rate? Arista looks more like Force 10 than Juniper. A story of good technology and bad execution.

    Share
    1. Given that even Juniper took a while to hit its stride, I wouldn’t write them off as quickly. The trends have finally converged for them and so has the market. It might have taken them a long time but they are growing really fast and are boosting their profits.

      Share
  4. Every other company that produces (or assembles in case of Arista) networking products think they are competing with Cisco, but in reality they are helping Cisco grow big time, by identifying the gaps. There is nothing in this world that Cisco can’t produce, and in time. Cisco has done it regularly and has done better. Arista shall either be a history or an acquisition in coming future.

    Share
    1. Hari

      Fair points. However Cisco is trying to deal with many markets and problems. I think there is an opportunity for its rivals to standout and it all depends on Cisco’s ability to move fast and strike back.

      Let’s see how it all shakes out.

      Share
    2. Hari, not sure I understand your point of producing vs assembling? The Cisco Nexus 3064 is a merchant silicon system, as were many of the Catalyst 29xx series as are the Cisco ISRs. The Arista portfolio is based on similar merchant silicon.

      The innate value though, to all of us who use network products is that we use the software, the hardware the accelerates some function the software enables. From interacting with Arista (I am not an employee of Arista or any other networking vendor) I have seen that they are far more focused on developing their software – and this is the key differentiator, not on feature depth – but on feature velocity enabled by a coherent and modern software architecture.

      Cisco’s OS architecture is decades old – this was made painfully obvious to me a few weeks ago when entering one command ‘show vlan’ crashed the entire Nexus 7000 – brilliant. Clearly this is an endemic problem of massively distributed uncoordinated teams with no clear leadership and no common architecture. Cisco was a great company in the 90s, and it seems the model that made them great is the model Arista is following today.

      Alan

      Share
  5. Find us on Facebook for even more about PLAY Activity Center! http://t.co/UmdvqGtz

    Share

Comments have been disabled for this post