Web Constructor 3.8 Performance is 64.3% higher than that of Web Constructor 3.7.1

April 6, 2006

Since Web Constructor 3.8, in addition to block cashing, supports page cashing we decided to check its impact on Web site performance.

Testing

For load testing we generated two domains with similar content (content of our esector.ru URL) — v371.esector.ru (with Web Constructor 3.7.1) and v38.esector.ru (with Web Constructor 3.8).

Tested were 5 possible configurations:

  1. v371.esector.ru without cashing
  2. v371.esector.ru with block cashing
  3. v38.esector.ru without cashing
  4. v38.esector.ru with block casing
  5. v38.esector.ru with block and page cashing

Apache JMeter, software for load testing of different Web applications, simulated 30 parallel Web site users and imitated their behavior.

For telling Web Constructor performance only, the client was not to request pictures and flash.

Testing was conducted on a regular lap-top rather than a server machine.

Those interested in details could see testing scenarios in the following documents esector-ru-v38.jmx (62.6 ÊÂ) and esector-ru-v371.jmx (62.6 ÊÂ)

Testing Results

Testing allowed us to get the data, as shown in the diagram, enabling us to conclude that block and page cashing improved Web site performance by 64.3 %.

Web Constructor performance does not however mean Web site performance. To bring two performance parameters to a common ratio, a channel is needed with no less than 5.87 Mbit/s capacity.

Unfortunately we have not found any performance data of other content management systems, therefore we could not compare them against Web Constructor 3.8.

We assume that performance of some content management systems may be on a higher side. Nevertheless, they do not enjoy the functionalities of Web Constructor 3.8

This is not yet the limit …

Performance could have been higher if:

  • Zend Optimizer, eAccelerator and ionCube PHP Accelerator had been used. For instance, when eAccelerator is used, Web site performance increases 3-4 times (up to 190-250 requests/sec);
  • Server machine had been tested; and
  • Fine tuning of server software had been used to maximum extent.