Has Cisco gone too far?

Miercom testing once again comes in to quesion.

Last week I read this blog on the Riverbed website:

http://blog.riverbed.com/2009/09/thank-you-cisco-thank-you-miercom.html

If true, the events portrayed in this blog are quite interesting.  Miercom, an "independent" research company commissioned by Cisco, claims that Riverbed's Steelhead 6020 appliance exhibited "system instability" when exposed to ~3000 TCP connections.  They used official-looking reports to lead readers to that conclusion.  Miercom's report can be accessed at:

http://www.miercom.com/dl.html?fid=20090915&type=report

The following is an excerpt from the Miercom report:

The test results shown in Figure 1 on page 1, illustrate that Cisco WAAS appliance achieved full target 4,000 concurrent sessions, while the Riverbed RiOS Steelhead appliance exhibited system instability and process crash at 3,012 concurrent sessions. In a three minute period of complete TCP blocking, the Riverbed Steelhead appliance took an additional 16 minutes to recover before TCP optimization resumed, as the device completed the system dump of log information. This recovery time could vary between 30 to 45 minutes as illustrated in Riverbed technical notes.

These are very profound allegations, and at first I didn't know who to believe.  The so-called "independent" 3rd-party who supposedly provides a neutral perspective?  Or Riverbed, the leading vendor in WAN optimization.  So I did my own research, by directly asking one of Riverbed's customers.  The following is a screenshot from the management interface of their Steelhead 6020 that they provided to me:

According to the customer, the above diagrams show a Riverbed Steelhead 6020 delivering optimized performance for more than 25,000 TCP connections.  Assuming each user consumes on average about five TCP connections, this represents a user population of about 5000 remote users that are receiving optimized performance through this particular Steelhead 6020.  It seems from this data that Miercom's assertion that the 6020 exhibit "system instability" at ~3000 TCP connections is clearly wrong.  Note that the above shows the management console of a Steelhead that is in an actual live production network environment, as compared to Miercom's test which was performed in a lab environment.

The customer also told me that he was extremely happy with his Steelhead 6020's.  He even offered me the following data on the performance benefits that they received from this 6020:

So the above findings bring me back to the events that are portrayed in the Riverbed blog.  Certainly, the findings from my research seem to support Riverbed's assertion that Miercom and Cisco rigged their test results unfairly.  That leaves me wondering, has Cisco's obsession with Riverbed driven them to resort to borderline ethical means in a desperate attempt to bring them down?  Was Miercom just a hired gun paid by Cisco to perform the hatchet job on their smaller competitor with the superior product?

• Cisco WAAS
• Larry Chaffin
• Miercom Test Labs
• Pluto Networks
• rievrbed steelhead