Arista v. Cisco: No Comment
April 28, 2010
A week ago, Arista Networks introduced its 7500 series server. In the process, founder and famed hardware innovator Andy Bechtolsheim took square aim at rival Cisco Systems. In public.
At the 2010 High Performance Computing Linux Financial Markets Show and Conference at the Roosevelt Hotel in New York, he threw up on screen for all attendees of the keynote session to see a head-to-head comparison of the specs of the 7500 with “five times the performance” at a “breakthrough price.”
The comparison was with the Cisco Nexus 7010. To get 384 wirespeed ports capable of shipping 5.76 billion packets of data a second, you’d have to have nine full racks of Cisco servers at a list price of $13.7 million, he said. The same capability would take just one-fourth of a rack with Arista. At $400,000.
This kind of frontal pillorying in a public forum does not sit easily if you’re the one being pilloried. A couple sessions later, Cisco’s global senior director of financial services Paul Jameson felt compelled to intone that “this becomes kind of a giant he-said she-said” conversation. And that Cisco had spent a half-million dollars to close down that kind of discussion, through “real world testing” using independent third parties on performance of an entire network architecture. That testing used “real world data” to compare performance, apples to apples.
His contention: That Arista’s gear does not stack up that well when dealing with “real world” data spikes and conditions. That it was prone to losing data under “burst” conditions – which happen when stock markets suddenly get active.
Now, he did not use Arista, by name. In public.
But his implications and contentions were clear. And it was only a few steps away from the dais to the Cisco suite – where the comparison was being carried out in private. With real servers.
In a single cabinet that could be mistaken in other rooms in a nice hotel as a 64-bottle wine cabinet sat a single Arista server and a single Cisco server, each attached to an Ixia traffic generator.
The point: To show first-hand to any potential customers who wanted to listen just how a Cisco Nexus server stacks up to an Arista rival.
Also being handed out were the results of a “real-world” set of tests comparing the performance of the Cisco Nexus 5010 and 5020 with that of the Arista 7148SX and 7124S.
“Detailed test results follow and demonstrate the advantages in using the Cisco Nexus 5000 in a network environment that consists of high, bursty traffic,’’ said Rob Smithers, the chief executive officer of Miercom which produced the lab report.
But the most interesting part of the Executive Summary was the first line. It said:
“Cisco commissioned Miercom to conduct an independent third-party performance test” of the Cisco and Arista switches.
Come again? Cisco commissioned the test? As an “independent” evaluation of performance?
That’s like Hershey commissioning a test of chocolate that proves it doesn’t cause zits. Independent? Third-party? Not if you’re the outfit who commissioned it.
