Riverbed Technology Inc. next week will jump into the application switching fray when it introduces its own unique twist on the function—a fusion of compression, caching and Web acceleration.
The San Francisco startup chose to address more than just slow Web application response times for traffic traversing a WAN with its new appliances, which combine multiple functions to speed TCP application response times.
The Steelhead appliances are targeted at distributed organizations with multiple remote locations. Installed at both the data center and remote offices, the appliances recognize “chatty” applications and the redundancies in the protocol handshakes that go on between TCP client and server, and then eliminate the redundancies to boost performance across the WAN by 100 times, according to Alan Saldich, marketing director at Riverbed.
“For things like (large) file sharing, what normally takes eight hours can be sent in 10 minutes. If you have a 100MB AutoCAD file, its like emptying a bucket with a teaspoon,” he claimed.
For applications that use TCP-based protocols such as the Windows Common Internet File System or Messaging Application Programming Interface, a simple action such as dragging a file from a remote file share to a local desktop generates 3000 or 4000 client/server interactions. And those application protocols will limit how much data in the payload can be sent at one time to between 8K and 64K.
The appliances reduce that by making “predictions about whats about to happen and minimize the round trips that go back and forth without impacting the application. They can keep from sending data thats already been sent,” he described.
“We say, weve seen this protocol, we know youre going to read the whole file. Well predict those 4000 interactions that will take place and well issue those client requests on your behalf on the server side—package them up and send them over the WAN as one or two responses,” added Eric Wolford, vice president of marketing at the company.
The appliances also act as a TCP proxy to terminate TCP sessions across the WAN. They include a hard disk that can range in size from 40GB to 512GB, which is used to store WAN traffic that goes through the appliance. The appliances monitor traffic flowing through them and when they spot data that has already been sent, they eliminate the redundancy. “Whether the file name changed or the method of sending it changed, our boxes recognize any redundancy and wont send it again,” he explained.
That optimization function, dubbed Scaleable Data Referencing, is distinct from the next-generation compression algorithms used by competitors such as Peribit Networks Inc. and Expand Networks. “Its suppression rather than compression. We store things on disk at both ends and reference it—sort of like a caching mechanism—but we fuse compression, caching and protocol acceleration,” said Wolford.
The appliances, available now in five different models, start at $5,995.