Dec 12, 2011
Speaking of technology, I can’t help but be amused at Riverbed’s brand new tunnel-based architecture. Aside from the fact that a new architecture that hasn’t been proven in the marketplace is a challenge in and of itself, this is particularly curious since it represents a complete U-turn and reversal of Riverbed’s strategy.
Riverbed now claims limited support for compressing some (not all) UDP applications by tunneling it in IPv4. Interesting, since for years the company has been trashing tunnels as “too technical” and dismissing UDP as an insignificantly small market segment. All the more reason to suspect that this new announcement is more a marketing ploy on Riverbed’s part than a serious attempt to provide a new technology solution.
By contrast, Silver Peak had the vision and foresight to support non-TCP traffic from the beginning. Silver Peak is proven through countless installations that demonstrate ease-of-deployment, scalability and extreme robustness. Not to mention our multiple patents in this space and over six years of experience installing, maintaining and improving that technology. Frankly, the barrier-to-entry has been set extremely high, and our leadership in this area would be extremely hard to match.
After all, packet mode optimization is not a minor tweak to RiOS code; it is a completely new approach. Even if Riverbed was serious, the Silver Peak VXOA software includes key features, such as auto-optimization, auto setup, cross flow compression, and loss mitigation capabilities that would be difficult, if not impossible, for Riverbed to implement without infringing on our intellectual property.
So, what does this really mean?
From a marketing standpoint, this new feature appears to be a clear and direct response to the fact that Silver Peak is the recognized leader for optimizing all IP traffic.
From a technology point of view, well, frankly, we’re surprised. First, Riverbed is only focusing on “deduplication” and “UDP.” Even a serious challenger should be able to support all non-TCP traffic, not just UDP. Clearly, Riverbed is not attempting to optimize GRE encapsulated traffic, which is essential for data center bridging solutions like Cisco OTV (for the record, Silver Peak optimizes all FCIP with no plugins and we’ve optimized IP in IP). We have also supported IPv6 encapsulation in IPv4 for several years.
In the company’s announcement, Riverbed only lists a few tangential UDP-based replication products, such as Aspera and VVR. What about VoIP? Video? VDI? EMC VPLEX? Many of these solutions run over UDP, too. However, they were conspicuously absent in Riverbed’s announcement. I have to ask myself: how serious is this move?
And, to be effective, Riverbed must go beyond bandwidth to perform dedupe, loss correction, out of order packet correction, QoS, and traffic shaping on all IP traffic — stuff that comes standard with Silver Peak. That’s the only way to truly have a positive affect on replication, voice, video, and a wide range of non-TCP applications where bandwidth is not the primary challenge.
While it’s great to see Riverbed jumping on the train for optimizing non-TCP traffic, and it is somewhat flattering that Riverbed is attempting to imitate our approach, Riverbed has a long way to go if it wants to pose a serious challenge to our ability to optimize all IP traffic across an entire enterprise.
Does the Riverbed marketing department have that kind of stamina? We shall see.