• United States
by Steve Taylor and Joanie Wexler

Using traffic shaping to improve response time

Dec 04, 20033 mins

* Where compression leaves off and shaping begins

Last time, we discussed some ways compression techniques can help achieve greater throughput for file transfers, provided the data being transferred is inherently compressible. But the link between compression and application response time is not well established. And in some cases, it’s nonexistent.

Compression applies algorithms to chunks of data to make the transfer more efficient. And if your application is bound up because a large data chunk is needed for each transaction, compression can be of help. However, this tends to be the exception rather than the rule, because most transactions involve only a relatively small amount of data per transaction.

With many application response problems, including voice over IP, the genesis of the problem is that other traffic, such as the file transfers that we’ve been talking about compressing, holds up delay-sensitive traffic. Whether the file transfer is compressed or not, it’s using bandwidth that is needed for the application that’s experiencing problems.

This is where traffic shaping comes in.

Note: The term “traffic shaping” can have a specific or a general meaning. Some use the term to refer to any number of quality-of-service mechanisms that “tame” traffic behavior, including prioritization and queuing, allocating a minimum amount of bandwidth to a particular application or protocol, or rate limiting traffic of a certain type. Some, however, use the term to describe pacing the rate at which packets enter the network to avoid packet loss, either because there is a speed mismatch at the sending and receiving ends or because the network is experiencing congestion.

For purposes of this newsletter, we’re using the first, more general, definition.

As we’ve often discussed here, traffic shaping is quickly moving up the OSI stack so you can even specify priorities based on the actual application. Think of this as a per-application committed information rate, possibly used in combination with a “priority permanent virtual circuit.” Each application is allocated some amount of bandwidth and/or a place in the pecking order.

To use a transportation analogy, compression gives you a much more efficient manner of packing a large truck full of widgets. But to transport these widgets you still need to get trucks on the highway. Even if there are fewer trucks, other traffic can still get stuck behind them.

Traffic shaping introduces express lanes and metering lights for the highways. The mission-critical traffic gets to go directly to the express lane, while the trucks may have to deal with admission control and contend with other non-express traffic.