OPAC Performance Improves

Submitted by palniadmin on Wed, 02/08/2006 - 11:48am.

Ex Libris recently profiled the PALNI system for OPAC performance bottlenecks, and found an Aleph V.17 problem that explains some of the slowness we've been seeing. Around January 26, they installed a temporary fix for the problem, and we have seen some improvement on our test keyword search. The test search is a simple, 2-word Boolean search that we've been running regularly over the past two years to monitor performance of the PALNI OPAC.

Since this fix was installed, average response time on our test search has improved significantly from over 5 seconds (between 1/16 and 1/22) to around 2.3 seconds (between 1/30 and 2/5). However, this is still about two times slower than we were seeing on the same search before our upgrade to Aleph V. 17.

Also, we've written software to extract actual response times from our web server log files for all OPAC keyword searches. That data shows that, between February 1 and today, Feb 8, average response time across all keyword searches against the PALNI OPAC was 4.6 secs. About 60% of keyword searches took less than 3 seconds, but more than 25% took more than 6 seconds.

We continue to have an open problem on this issue with Ex Libris. Following is the lastest PALNI OPAC performance update that we sent to Ex Libris on Feb. 8.

2/16/2006

We gave both good and bad news to report on PALNI's OPAC response time problems:

  1. Following the change to the OPAC software that Ex Libris installed sometime around 1/26, response time on our keyword test search (a 2-word Boolean search that we've been running regularly over the past couple of years to track OPAC performance) improved significantly.

    For example, before the change, for the 7 days from 1/16 through 1/22, average response time on this test search was 5.25 secs, with a peak of 22.34 secs and a minimum of 2.23 secs. After the change, for the 7 days from 1/30 through 2/5, ave response time on the same search dropped to 2.34 secs, with a peak of 7.07 secs and a minimum of 1.56 secs.

    So, response time has definitely improved as a result of the fix.
  2. However, response time on our test/benchmark search is still signficantly slower that we were seeing with Aleph V. 15.

    For example, a year ago with V. 15, for the 7 days from 1/30/2005 to 2/5/2005, ave response time on the same search was 1.27 secs, with a peak of 4.93 secs, and a minimum of 0.86 secs.

    So, on our test search, response times are still about two times slower with V. 17 compared to V. 15, even after the recent fix.
  3. We've also extracted web server response times for all keyword searches from our Apache log file. Since February 1, average response time for all keyword searches has been 4.60 secs. 27% of our keyword searches took more than 6 secs, and 17% took more than 10 secs.

In summary, after the recent fix, performance on our standard comparison search improved significantly. We appreciate your work to identify and fix this OPAC problem.

However, it it is still about 2 times slower than we were seeing with Aleph V. 15. Also, ave response times for all keyword searches logged by Apache are between 4 and 5 seconds, and PALNI libraries are still reporting that the system seems slower than they were seeing with Aleph V. 15.

Vince

printer friendly version | 207 reads
( categories: Aleph News and Information | OPAC )