Hi,
I have been doing some load testing on the development version of
DSpace3 here at Cambridge. (Actually I am doing it from Sydney where I
am based, but thats another story).

I've done some reasonable extensive JMeter load tests all of which
seem to point to quite slow performance for a web application. But I
can illustrate quickly with just Apache Benchmark.

First off, if I use Apache Benchmark on our live system running jspui
to fetch an item page with:

ab -k -c10 -n100 http://www.dspace.cam.ac.uk/handle/1810/244032

I get

Requests per second:    7.60 [#/sec] (mean)
Time per request:       1949.402 [ms] (mean)

The instance is DSpace 1.6 and unloaded other than the quick test. No
one has been complaining about the live instance and many are saying
its quite "nippy" even if the figures say otherwise.


If I do the the same against the new DSpace3 instance with no local
modifications other than I skinning I get.

Requests per second:    13.73 [#/sec] (mean)
Time per request:       728.264 [ms] (mean)

If I do the same for a custom thumbnail servlet using the DSpace3 API
I get with caching in the Apache Httpd fron end turned off

Requests per second:    873.10 [#/sec] (mean)
Time per request:       11.454 [ms] (mean)

If I turn caching on ant let the httpd front end deal with the request I get

Requests per second:    17525.41 [#/sec] (mean)
Time per request:       0.571 [ms] (mean)

If I attached  profiler and look into exactly what is happening.
1 request with the profiler connected takes 970ms
Some observations:
The statement cleanup from the SQL connection pool is about the most
expensive operation at 110ms for 57 invocations.
To build the page takes 57 SQL queries, although none are expensive.
(all bar 3 report < 1ms).
UsageActionLogger.act consumes 2/3 of the time spent in the request
and all of that is in SolrLogger.storeParents

Is < 20 request / s with a 10 concurrent request threads normal for DSpace ?

Is anyone running DSpace with a Apache Httpd front end with mod_cache
enabled, presumably forging statistics ?

Any insights would be helpful so I can advise the team here on what to do.

This is important to us as we are expecting an increase in usage of
the service due to the UK open access agenda and I need to get
performance up to more typical levels for anon users. (ie > 2K
pages/s). Authenticated users are ok with the current performance.


Sorry for the long email.
Ian

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
DSpace-tech mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dspace-tech
List Etiquette: https://wiki.duraspace.org/display/DSPACE/Mailing+List+Etiquette

Reply via email to