Arista Grows Up, Pumps Iron

Arista Networks Inc. wants to be the next Juniper Networks Inc. (NYSE: JNPR).

That is, the company wants to muscle in on Cisco Systems Inc. (Nasdaq: CSCO) in the data center market, just as Juniper did in the telco router market.

And its first real test starts now. After picking up more than 300 customers for its small, top-of-rack switches, Arista is trotting out the big iron: the Arista 7500, a 10-Tbit/s switch line being launched this week.

The 7500 is generally available now, but two revenue-generating customers have had it in operation for five months, says Douglas Gourlay, Arista's cultured vice president of marketing. "We didn't want this to be the Macbeth launch, full of sound and fury, signifying nothing," he says.

It's the biggest salvo yet from the startup founded by Sun Microsystems founder Andy Bechtolsheim and run by former Cisco executive Jayshree Ullal. (See Ullal Lands in the Cloud.)

And while Arista claims to be attracting customers with the benefits of modern chips and a new operating system, its achievements might have just as much to do with the names behind the company.

Arista is claiming double- and triple-digit sales increases per quarter, but then again, it's only been shipping gear since late 2008. Still, there's no question Arista has generated buzz.

"The way you would evaluate a company this young is to look at the interest around them -- and I get a lot of questions about them," says Zeus Kerravala, an analyst with Yankee Group Research Inc.

With its star power, the privately funded, VC-less startup was able to do things in reverse. Most switch/router startups introduce big-iron products first, to prove that the architecture works on a big scale. But building and shipping those prototypes eats up money in the millions of dollars, Gourlay says, as potential customers try the system in labs.

Arista started small, with pizza box (1 rack-unit), top-of-rack switches. They were relatively cheap to build and ship, and Arista claimed some noteworthy figures for high density and low latency. (See Arista Counts 10GEs.)

But Arista also got access to potential customers that a typical startup wouldn't enjoy, Gourlay admits.

And now that its small switches are deployed, Arista is bringing out the 7500 to aggregate them -- "bringing the product out in line with when the customers start to feel the pain of not having a cost-effective aggregation platform at that density," says Gourlay.

So far, the plan has worked. Every 7500 that's shipped has been for revenue, Gourlay says.

The data sheet has some eye-catching numbers, if nothing else. The 7500 has switching and per-slot capacity of about 10.3 Tbit/s. In terms of interfaces, the box starts out with a maximum of 384 10-Gbit/s ports for what might be called a 7.68-Tbit/s capacity; a demo at Interop will show the box running fully packed.

That capacity comes in one-fourth of an equipment rack and eats up 13.2 Watts per port, Arista claims.

Inspired by Juniper
Arista started out to build, not switches, but a new network operating system, one that would identically run on anybody's chips. It was meant as a Linux-based antidote to the numerous versions of IOS that Cisco runs on different product lines. Even Juniper, whose tight rein on JunOS serves as a model for Arista's software roadmap, isn't delivering exactly the same code to all JunOS boxes, notes Gourlay.

In a move reflecting Infinera Corp. (Nasdaq: INFN)'s chip-to-systems development, Arista decided in 2007 that the best way to bring the OS to market would be to build the system that used it. The company did more than take advantage of its operating system -- it took advantage of off-the-shelf chips and a backplane built from scratch, coming up with designs it says are more compact and less power-hungry than bigger companies'.

"Juniper had its growth when the largest routing incumbent had its eye off the routing ball and was growing the switching business," Gourlay says. "Right now, with disruption to their partnerships with a move into the server business, and chasing 40 or 50 market adjacencies with a lack of unity of demand, I think we end up in a position to say there's an opportunity." (Hint: He's talking about Cisco.)

In addition to Cisco, Arista will be battling data-center-minded companies such as Brocade Communications Systems Inc. (Nasdaq: BRCD), Extreme Networks Inc. (Nasdaq: EXTR), and Force10 Networks Inc. Router and switch vendors Juniper and Alcatel-Lucent (NYSE: ALU) could be considered competitors, too.

Naturally, the established data-center camp isn't willing to concede to Arista yet, especially considering the company is just starting to ship large switches.

"They might have a short-term competitive advantage on density, but we wouldn't expect that to last too long, especially using merchant silicon as they do," says Kevin Wade, senior director of marketing at Force10. "Historically it was always back and forth with Force10 and Foundry [now owned by Brocade], which we survived successfully. If there's someone else throwing their hat into the ring, I'd say I'm not concerned."

— Craig Matsumoto, West Coast Editor, Light Reading

Page 1 / 2   >   >>
StartUpGuy1 12/5/2012 | 4:39:18 PM
re: Arista Grows Up, Pumps Iron

Interesting comment from the guy at Force10...  Especially about the use of the merchant silicon in the Arista switch.   


Looks like a very cool box with a footprint and power usage that puts a lot of vendors in catch up mode..

DouglasGourlay 12/5/2012 | 4:39:12 PM
re: Arista Grows Up, Pumps Iron

Me cultured?  Why thank you Craig!  We have a couple webcasts going on Thursday for anyone with questions or wanting a bit of a deeper dive into the platform.


I agree with the sentiment that 'biggest. fastest, densest' is not the most long term and sustainable value proposition.  But couple that with an open O/S environment that has people designing into it, rather than around it - and life gets intriguing.  :)

"Once ore into the breach dear friends!"



Archinet 12/5/2012 | 4:39:11 PM
re: Arista Grows Up, Pumps Iron

I think that it was worth mentioning Voltaire and Blade Networks especially in the light of their partnership in providing extremely high density Ethernet switch fabrics

quicktime 12/5/2012 | 4:39:10 PM
re: Arista Grows Up, Pumps Iron

If carefully designed,  will this be a big hit to Cisco/Juniper in some market?

Lower cost, stable softwrae, fast product rolling, there are some edges there.

When SGI booms, will it laught at Nvidia?



Pete Baldwin 12/5/2012 | 4:39:09 PM
re: Arista Grows Up, Pumps Iron

For the longest time, chip and systems companies have said that the ASIC model at Cisco/Juniper/Alcalu/etc. won't be sustainable.  That may slowly be starting to come true, with Cisco and Juniper using EZchip, but it hasn't been enough to let anybody make a big breakthrough using merchant chips.

At the same time, Arista (& others) have a point about merchant chips: They can take advantage of more recent manufacturing processes to produce smaller, lower-power devices. 

Arista's hope is that that advantage propels them.  But really, it's the executive team that's key -- no matter how good their product might be, they wouldn't have gotten the chances they did if they were a no-name startup.

Pete Baldwin 12/5/2012 | 4:39:08 PM
re: Arista Grows Up, Pumps Iron

OK, they're now mentioned, thanks to you.  See, the system works!

How do Voltaire+Blade stack up against the biggest & baddest from , say, Brocade or Extreme or Force10?

Pete Baldwin 12/5/2012 | 4:39:08 PM
re: Arista Grows Up, Pumps Iron

More of a hit to Cisco, in absolute numbers.  Percentage-wise, it would be a hit to Juniper, since Juniper's enterprise sales are smaller but a key part of the company's growth plan.

Companies more dependent on the data center probably have more to lose than Juniper. That's why I bugged Force10 about this, when I got the chance.

Even if Arista succeeds, it's going to take a long time for it to make a dent, I'd think.

Pete Baldwin 12/5/2012 | 4:39:08 PM
re: Arista Grows Up, Pumps Iron

Thanks for posting, Doug.  I have to confess: It was our copy editors who added the word "cultured."  Me, I was just reporting you as "guy with iPad." :)  But I give you credit for getting the Shakespeare attribute right.

yes, the open OS environment is promising -- and to catch everyone else up on this, it's the idea of letting other developers write applications to the OS, just as the Linux world does.  Juniper is applying this idea, in controlled doses to companies like Blade.

Maybe that's where the longer-term promise is, but it sounds like a lot more work than, "Hi, we're Jayshree Ullal and Andy Bechtolsheim, and our Ethernet switch has better specs than yours."  We'll see.

quicktime 12/5/2012 | 4:39:06 PM
re: Arista Grows Up, Pumps Iron

Whether Arista Networks could stand out, I have a couple of questions:

1) Will ASIC-based system excel COTS-based system a lot in performance?

    If not, there are a lot of potentials there.

2) Will their system software be more stable and scalable? 

   For hardware-based companies,

   they focus more on hardware than software, but for customers, they are system.

3) The threshold for Ethernet/Switching market is become lower; Will the service 

provider adopt switch to such solution? They feel comfortable, and have confidence?

4) How about solutions from ZTE, Huawei?

    Though they are blocked out of North American market temporarily;

    what they adopted will form pressure to Cisco/Juniper.

    Innovation comes from competing.


As to Open System,  how could data network achieve telecom-level stability?

Open is a buzz word for marketing, everybody has its explanation. If means

API, I believe every vendor has such stuff. If means source code, I doubt that.


Just some random thougts



Pete Baldwin 12/5/2012 | 4:39:05 PM
re: Arista Grows Up, Pumps Iron

Very good questions.  To answer 3 and 4, I agree the Ethernet/switching threshold is lowering, and I think it's likely that low-cost Asian equipment will eventually (years from now) dominate that sector even within service providers, except at the very high end.

#1 is a key question and one the chip guys have wrestled with for so long.  Obviously Cisco, Juniper, AlcaLu all see merit in ASICs for routing, but for Ethernet, it does seem COTS is catching up.

Thanks for reminding us about the malleability of "open," too.  Cisco and Juniper each talk about open standards, but neither means "open" the way Vyatta does...

Page 1 / 2   >   >>
Sign In