Americas

  • United States
by Steve Taylor and Joanie Wexler

Compression and traffic shaping put to the test, Part 2

Opinion
Dec 11, 20032 mins
Networking

* Interactive Web sessions and HTTP file transfers with and without compression and traffic shaping

Today, we’re continuing our discussion of real-world testing of applications-aware processors that enhance the apparent throughput of mission-critical applications by using both compression and traffic shaping techniques.  The test of interest today was performed by the Tolly Group on behalf of Packeteer.

The test bed was very straightforward.  Two clients and two servers were used; one client and server ran an interactive Web session (via Citrix) and the others performed an HTTP file transfer.  Both applications ran across a shared simulated 64K bit/sec circuit.

The test began by measuring baseline values with no compression or traffic shaping applied.  The interactive Web session’s performance was judged as “poor,” and this was established as a baseline for the file transfer time.  When compression was added into the mix, the file transfer time improved by about 35%, but the interactive Web session’s performance remained at a “poor” level.

Not surprisingly, when traffic shaping was added to compression, the performance of the interactive application rose to a “good’ status, and file transfer time also rose to almost a 50% performance increase. 

The part of the test that was a little surprising, though, was when traffic shaping without compression was used.  In this case, as expected, the interactive application rose to good performance.  One would think, however, that the performance would be at the expense of the file transfer.

On the contrary, the testers saw “good” interactive performance and recorded an increase of almost 25% in the performance of the file transfer.  On closer examination, this improvement was attributed to Packeteer’s traffic shaping process called “TCP rate control,” which influences the generic TCP flow control behavior.  As such, the TCP session allocated to the file transfer has a much more realistic view of how much bandwidth is available and less time is spent increasing and decreasing the flow control parameters.

If you want more details of this test, please see the link below.