posted by: Eric Siegel
I'm back from Interop, and I must admit that every year I think, "This is the year! People will finally understand how useful these WAN performance optimizers can be." Well, I've now been saying that for four years; it's time to admit that things are going more slowly than I would like.
The developments this year at Interop were more evolutionary than revolutionary; in addition to simplifying management and increasing scalability in both the number of devices and their capacity, more vendors are providing software clients for user laptops. Those clients let remote users or small offices accelerate TCP flows and accelerate access to remote file shares using Microsoft's CIFS file access protocol. (Other technologies, implemented in application front end processors, have done that for browsers for years.) Orbital Data (now a part of Citrix) was an early provider with what is now called WANScaler Client; both Blue Coat and Riverbed are providing their own software clients (ProxyClient and Steelhead Mobile); and now Cisco (WAAS Mobile) and Nortel (Mobile Client Accelerator) have entered the market using externally-developed software.
An additional advantage of software clients is that they allow the optimization of encrypted data flows, because they can see the flow before it's encrypted by the laptop for transmission. Alternatives for optimization of encrypted flows rely on giving the flows' private keys to the optimization system. That approach, pioneered by Certeon, is now evolving and being implemented by other vendors as well. (Blue Coat, Certeon, Cisco, Juniper, and Riverbed have implemented various methods.)
I think that a major reason that people are cautious about optimizers is a side effect of how effective they can be and how much money they can save. A lot of vendors have been attracted to the market, and one way that they differentiate themselves is by telling potential customers how complex and confusing their competitors' systems are. Yes, integration of optimizers can be complex, but that's no reason to avoid investigating them! For many architectures, integration is straightforward, and the benefits are huge. They really can provide both stunning improvements in performance at the same time that they cut costs.
So jump in now; don't waste years in a slow evaluation process. Just remember to set expectations properly: you will probably want to re-evaluate and re-engineer this solution in three years. But it will probably pay for itself, in hard savings, long before then.
Comments