In March 2000, Polyteam benchmarked DynaCache-30i, a caching proxy offered by InfoLibria. The tests were executed in PolyLab using the PolyMix-2 workload.

We have followed the second bake-off rules and methodology to ease comparison of the results. This short report does not describe the workload or the caveats of interpreting the results. Familiarity with the official bake-off report is a must!

We also note that while the workload for this test is identical to what we used during the second bake-off, it is somewhat unfair to compare competitors results at the bake-off with the results achieved in a lab: Bake-off participants must meet the bake-off deadlines while a lab test is scheduled according to vendor preferences. Also, other vendors may have improved the performance of their products since the bake-off but did not officially test their new products yet.

We suggest that you use these results primarily as an indication of the progress done at InfoLibria, rather than for inter-vendor comparison with bake-off figures.

As always, we do not give purchasing advise. We provide performance measurements and expect the reader to use this information, along with other important factors, in their decision making process.

Table of Contents

1. Where to find more information
2. Executive Summary
    2.1 Performance Changes since January
3. Product Configurations
4. Comments
    4.1 Polyteam Comments
    4.2 Vendor Comments

Only major performance measurements are highlighted in this document. For more information, consider these sources:

• Auto-generated individual reports with raw measurements and trace data.
• Auto-generated tables and charts comparing the two products.
• Detailed configuration information including bake-off tuning options.
• Complete Polygraph logs.
• The official bake-off report where January InfoLibria results were borrowed from.
• Web Polygraph site with documentation, sources, independent test results, and other benchmarking info.

The ``Executive Summary'' table below summarizes the performance results.

Links in column headers point to bar charts comparing the corresponding measurement. Product links point to a auto-generated individual reports for the entries.

All published performance tests finished with less than 0.2% of failed transactions. Note that the rules disqualify a run with more than 3.0% of errors.

Overall, DynaCache-30i tested in March showed better performance. Here is a summary of major changes in DynaCache performance since it was tested in January.

• Positive changes:

  1. Mean response time went down from 3.1 sec to 2.0 sec.
  2. Hit response time went down from 1.6 sec to 0.4 sec.
  3. Document Hit ratio went up from 32% to 38%.

• Negative changes:

  1. Response time has higher fluctuations (compare January and March traces);
     Document hit ratio has higher fluctuations (compare January and March traces).
  2. Time until first hit went up from 3.2 min to 4.8 min.

Here are the configuration details for all tested products. The vendor used the same hardware configuration, but different software version and switching gear.

All numbers are totals for the entire tested configuration unless noted otherwise. See individual product pages (linked from the first column) for more details about box configurations.

Both InfoLibria products were tested in a ``transparent proxy'' setup.

None.

http://www.infolibria.com/

InfoLibria would like to thank the Polyteam for the opportunity to demonstrate significant improvements in the response time and hit ratio achieved by InfoLibria 30i. The software improvements made to this product are applicable to other DynaCache models as well and they will be applied to future releases. The periodic fluctuations in hit ratio and response time can be attributed to periodic garbage collection of the cache. An improved garbage collection algorithm is being implemented.