Opinion: Infrastructure's Need For Speed
A lot of those demands come from Web-based applications that drive a ton of audio and video. In fact, to circumvent constricted corporate networks that typically have severe limits on the size of files that can be sent, it's not unusual to hear about corporate users opting to use insecure public FTP sites such as Yousendit.com to send files to fellow employees, as well as to colleagues they are working with in other organizations.
That kind of behavior tends to make the typical risk compliance officer blanch, but until recently there really wasn't any effective way to send large files over FTP unless a company had some sort of point-to-point caching device set up between two specific locations.
This is why a startup called FastSoft is creating a stir in the world of networking; it claims to have delivered a solution for dramatically accelerating transfers over a Fast TCP transport using an appliance that only needs to be deployed at the sender's site. Developed by Prof. Steven Low and his team at the California Institute of Technology's Netlab, Fast TCP does a better job of managing the queues of packets that are in the network at any given time.
This has allowed customers such as The Post Group, a production company in Los Angeles, to transfer 500MB files in 20 seconds, according to CIO Darin Harris. This is a key capability, Harris says, because the nature of the company's business relationships with different clients varies from project to project. That makes it impractical to set up point-to-point caching appliances with every customer or potential business partner, so up to now the only practical alternative has been to burn a video on a disk and ship it to the customer via a messenger or overnight delivery service. The FastSoft approach solves a significant problem, Harris explains, because the application-level acceleration devices Post looked at typically don't support video or audio files, and proprietary content-delivery network services were too costly.
FastSoft is just one way I.T. organizations are working around what is becoming a desperate problem. For example, the advent of a device such as the FastSoft accelerator doesn't obviate the need for point-to-point application acceleration devices. In fact, some of these tools, such as the one provided by Blue Coat Systems, have evolved to the point where they only need "virtual devices" on one end of the connection because the receiving part of the application acceleration function is now handled by a piece of downloadable software. And, for the record, Blue Coat claims to support video.
Part of the problem that companies such as Silver Peak Systems, Riverbed Technology, Citrix, Coyote Point Systems, F5 Networks, Packeteer, Stampede Technologies, Juniper and Cisco are all trying to solve is the simple fact that the wide area network bandwidth available for most applications is only a fraction of the speed those same applications have come to expect on a local area network. And as these applications have been extended out across the Web, I.T. organizations are wrestling with how to maintain performance, short of refreshing the entire network infrastructure at what most organizations would call an unacceptable cost. But point-to-point application acceleration devices are not inexpensive, either.
The general expectation is that over time, Cisco and Juniper will embed the necessary level of application acceleration intelligence needed by their customers into the core routing and switching fabric without requiring customers to have dedicated devices to perform that task. But it looks like we're still years away from that actually happening to the number of application acceleration devices on the network, which in some ways are analogous to pacemakers; these should continue to expand at a fairly rapid rate to help compensate for the overall sluggishness of the network.
Unfortunately for a lot of I.T. departments, the business community has ambitions for extending business processes across the Web that typically fail to take into account the fragility of the network infrastructure. And because that infrastructure isn't likely to get better anytime soon, I.T. departments will have to come up with some creative approaches to network surgery just to keep the patient alive and breathing, never mind running at top speed.