On 6/7/12 9:37 AM, Lance Speelmon wrote: ... > filterCache > Concurrent LRU Cache(maxSize=512, initialSize=512, minSize=460, acceptableSize=486, cleanupThread=false) > <stat name="cumulative_hitratio" > > 0.65 ...
I think you should be getting a lot more benefit from the filterCache. (Some of the planned OAE 1.4 improvements rely on the filter cache being configured to scale.) For comparison, here's the Before of my simple short exploratory test: * queryResultCache LRU Cache(maxSize=512, initialSize=512) - cumulative_lookups = 8268 - cumulative_hits = 1017 - cumulative_hitratio = 0.12 - cumulative_inserts = 7240 - cumulative_evictions = 5692 * documentCache LRU Cache(maxSize=512, initialSize=512) - cumulative_lookups = 70266 - cumulative_hits = 57034 - cumulative_hitratio = 0.81 - cumulative_inserts = 13232 - cumulative_evictions = 11459 * fieldValueCache Concurrent LRU Cache(maxSize=10000, initialSize=10, minSize=9000, acceptableSize=9500, cleanupThread=false) - cumulative_lookups = 2602 - cumulative_hits = 2594 - cumulative_hitratio = 0.99 - cumulative_inserts = 4 - cumulative_evictions = 0 * filterCache Concurrent LRU Cache(maxSize=512, initialSize=512, minSize=460, acceptableSize=486, cleanupThread=false) - cumulative_lookups = 12525 - cumulative_hits = 12460 - cumulative_hitratio = 0.99 - cumulative_inserts = 65 - cumulative_evictions = 0 And here's my After, with the 25% reduction in mean response time: * queryResultCache - LRU Cache(maxSize=8192, initialSize=512) - cumulative_lookups = 6170 - cumulative_hits = 4865 - cumulative_hitratio = 0.78 - cumulative_inserts = 1298 - cumulative_evictions = 0 * documentCache - LRU Cache(maxSize=2048, initialSize=512) - cumulative_lookups = 66157 - cumulative_hits = 63255 - cumulative_hitratio = 0.95 - cumulative_inserts = 2902 - cumulative_evictions = 0 * fieldValueCache - Concurrent LRU Cache(maxSize=10000, initialSize=10, minSize=9000, acceptableSize=9500, cleanupThread=false) - cumulative_lookups = 2513 - cumulative_hits = 2501 - cumulative_hitratio = 0.99 - cumulative_inserts = 6 - cumulative_evictions = 0 * filterCache - Concurrent LRU Cache(maxSize=2048, initialSize=512, minSize=1843, acceptableSize=1945, cleanupThread=false) - cumulative_lookups = 9581 - cumulative_hits = 9419 - cumulative_hitratio = 0.98 - cumulative_inserts = 162 - cumulative_evictions = 0 These settings are still much lower than I see in online examples, but they're enough to show a promising trend. If you can get your filter-cache hit ratio up, I imagine you'd see more dramatic improvements than I did. Best, Ray > > On Jun 7, 2012, at 8:56 AM, Ray Davis wrote: > >> (Apologies in advance if this already in the plan -- I didn't spot it in >> a cursory search.) >> >> Can we make sure that load-test / post-bug-bash analysis includes the >> cache section of the Solr server's stats report? By default, this should >> be available from the server's host system at: >> >> curl http://localhost:8983/solr/admin/stats.jsp >> >> Cache adjustments are one of the most straightforward ways to improve >> Solr speed, and I think our current solrconfig.xml defaults may be a bit >> off. Yesterday I *finally* got a chance to run a quick exploratory Tsung >> test against our production data, and on my first try I got a 25% >> reduction in mean response times. >> >> Any data points available from other deployers? >> >> Best, >> Ray _______________________________________________ oae-dev mailing list [email protected] http://collab.sakaiproject.org/mailman/listinfo/oae-dev
