2.5G Carriers Look to Optimize Assets
One solution to this problem is to deploy network optimization software that can compress the volume of data sent over a wireless link. A host of startup vendors, such as Speedwise, Bytemobile Inc., First Hop Ltd., and Broadcloud Communications Inc., have sprung up and are bidding to supply this software to network operators.
Elaine Roberts, senior manager of service enabling technologies at Vodafone PLC in the U.K., which deployed Bytemobile’s solution to support the launch of a general packet radio service (GPRS) Internet Access Service last October, says that the benefits of optimization software are twofold: Customers get more for their money and network bandwidth is used more efficiently.
According to Daphna Bahat, VP of product marketing at Speedwise (a supplier to carriers such as Voicestream and Telecom Italia Mobile), data compression can improve performance by up to five times. For applications such as Web browsing, she says, compression works not only to reformat images to a lower resolution, but also to minimize the “chatty” nature of HTTP traffic by predicting requests to a Web server and then bundling information to the user. And, with most GPRS implementations offering real data rates of between 20 kbit/s and 30 kbit/s, this can have a huge impact on service.
Karl Freter, chief architect at Bytemobile, believes they have now convinced operators that optimization is essential, but he stresses that the technology should be transparent to the network. It is important to be able to manipulate data without being seen and to be able to scale independently of GGSNs (GPRS Gateway Support Nodes) because, says Freter, this cuts the time needed for interoperability testing prior to deployment. “As far as the client device is concerned, we don’t exist,” he adds.
Indeed, in the current climate, where carriers are happy just to get services up and running, the focus is overwhelmingly on client-free solutions, which are easier to implement and run with a standard HTTP browser. But, looking further ahead, Freter believes that “thin clients are going to make a significant difference” and that Bytemobile is already working on such a system with an as yet unnamed carrier.
However, the client/server approach, which notionally offers better performance and a greater opportunity for customized branding on end-user devices, is taking longer to get established.
BlueKite Inc., the first optimization player to win major customers, took the client/server approach. The San Francisco-based company developed a sophisticated HTTP optimization and caching system with client software that sits under the standard browser and servers on an operator's network. Despite the customer wins, Bluekite recently pulled out of the optimization business and shut its European office.
Operators also need more than just application-layer compression from an optimization product.
At the transport layer, GPRS and 1xRTT networks are characterized by high-latency and suffer from random packet loss of between 1 and 10 percent. And because TCP, which was designed for the fixed Internet, interprets all packet loss as congestion, data sessions suffer frequent retransmit timeouts and choking of the data steam, when in fact there is no problem on the link. The result is that TCP grossly underutilizes limited wireless bandwidth by between 5 and 40 percent.
Because of this, says Freter, optimization software should be able to determine if packet loss is due to congestion or wireless-specific issues and then massage the TCP stream between client and server to minimize the impact on end users. This can be done through a variety of techniques such as statistical differentiation of congestion and non-congestion loss, early response to congestion, caching, and coordination of rate management across concurrent user sessions. “It’s nice to compress data, but if TCP continuously restarts, it doesn’t really buy you much,” he explains, despite his assessment that TCP optimization accounts for only 10 to 20 percent of the value of Bytemobile’s wireless optimization service node.
The sales cycle for an optimization system is typically between six and nine months, with all vendors stressing easy scaleability to meet the anticipated surge in wireless data subscribers.
Optimization vendors tend to be very cagey about the pricing of their systems. But recently we've heard they cost in the ballpark of around $1 million for 25,000 to 50,000 users.
However, the battle between these young vendors is extremely fierce. And we've heard that recently vendors having been virtually giving away initial licenses in their efforts to penetrate the major accounts.
— Gabriel Brown, special to Unstrung