More tests
A discussion list for the collaboration between MassGIS, the GIS department for the state of Massachusetts, and OpenGeo, the geospatial division of The Open Planning Project.
Back to topic listing
Previous thread
Next thread
-
From: aleda
Subject: More tests
Date: Nov 30, 2009 03:39 PM
The NON tiled tests are taking a loooong time, so here's the other stuff meanwhile. http://maps.massgis.state.ma.us/temp/geoserver_raster_tests/tests_with_j ava_1_6_1.3G_mem/arcgis_server_to_maps_geoserver_11_30_09/ I got throught the tiled 1/5/10/15/20/25 and the non tiled 1/5/10/15, I'm running the 20 now. This is with tomcat 5.5.28 set_env.sh: JAVA_HOME=/usr/local/apps/jdk1.6.0_16 JAVA_OPTS="$JAVA_OPTS -Xmx1324m -XX:MaxPermSize=256m" IMG_USGSQUADM (with the projection fix) jpeg response format default logging some notes: tiled 1 user - done, no errors 5 users - done, no errors, maps top load up to 32 10 users - done, no errors, maps top load up to 49 15 users - done, no errors, maps top load up to 79 20 users - few timed out errors - connection timed out - maps top load up to 78 25 users - more connection timed out errors - maps top load up to 73 no errors in geoserver.log non tiled 1 user - done, no errors, maps top load up to 1 5 users - done, no errors - maps top load up to 4 - got kind of slow 10 users - done, no errors , maps top load up to 9 - got kind of slow 12:46:36 up 35 days, 6:17, 5 users, load average: 9.66, 9.28, 8.39 492 processes: 481 sleeping, 11 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 89.2% 0.0% 0.5% 0.0% 0.0% 0.7% 9.2% cpu00 90.1% 0.0% 0.9% 0.0% 0.1% 1.5% 7.1% cpu01 88.3% 0.0% 0.1% 0.0% 0.0% 0.0% 11.4% Mem: 3034532k av, 2973636k used, 60896k free, 0k shrd, 122020k buff 2317924k actv, 404936k in_d, 48932k in_c Swap: 9502400k av, 507784k used, 8994616k free 569560k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 3223 afreeman 20 0 1558M 1.5G 10588 R 20.8 51.4 1:41 1 java 2949 afreeman 24 0 1558M 1.5G 10588 R 11.8 51.4 2:44 1 java 2876 afreeman 20 0 1558M 1.5G 10588 R 10.7 51.4 2:38 1 java 3301 afreeman 16 0 1558M 1.5G 10588 R 9.8 51.4 3:05 0 java 1758 afreeman 15 0 1558M 1.5G 10588 S 9.6 51.4 1:55 0 java 3224 afreeman 24 0 1558M 1.5G 10588 R 5.5 51.4 4:13 0 java 3266 afreeman 24 0 1558M 1.5G 10588 R 5.4 51.4 2:01 0 java 2292 afreeman 16 0 1558M 1.5G 10588 R 4.9 51.4 3:34 1 java 2846 afreeman 25 0 1558M 1.5G 10588 S 4.5 51.4 3:16 0 java 2797 afreeman 25 0 1558M 1.5G 10588 S 1.9 51.4 3:50 1 java 1757 afreeman 16 0 1558M 1.5G 10588 S 1.1 51.4 1:09 1 java 1756 afreeman 16 0 1558M 1.5G 10588 S 0.9 51.4 1:11 1 java 1858 afreeman 15 0 1558M 1.5G 10588 S 0.4 51.4 0:05 0 java 1909 afreeman 16 0 2332 2260 804 R 0.3 0.0 1:01 0 top 1947 afreeman 15 0 1558M 1.5G 10588 S 0.2 51.4 0:06 0 java 32063 jbosstom 15 0 257M 150M 9144 S 0.1 5.0 0:10 1 java 1854 afreeman 15 0 1558M 1.5G 10588 S 0.1 51.4 0:05 0 java 1 root 15 0 352 304 268 S 0.0 0.0 0:15 0 init 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1Attachments
-
From: groldan
Subject: Re: More tests
Date: Nov 30, 2009 03:47 PM
Those are too bad results. I'll come back with a version that rocks on arcy, then we can run there and even profile as we did with Saul in the past. Gabriel. PS: I either find what's going on there or... well, I'll find out. Freeman, Aleda (EEA) wrote: > The NON tiled tests are taking a loooong time, so here's the other stuff > meanwhile. > _http://maps.massgis.state.ma.us/temp/geoserver_raster_tests/tests_with_java_1_6_1.3G_mem/arcgis_server_to_maps_geoserver_11_30_09/_ > > I got throught the tiled 1/5/10/15/20/25 and the non tiled 1/5/10/15, > I'm running the 20 now. > > This is with tomcat 5.5.28 > set_env.sh: > JAVA_HOME=/usr/local/apps/jdk1.6.0_16 > JAVA_OPTS="$JAVA_OPTS -Xmx1324m -XX:MaxPermSize=256m" > > IMG_USGSQUADM (with the projection fix) > > jpeg response format > > default logging > > some notes: > > tiled > 1 user - done, no errors > 5 users - done, no errors, maps top load up to 32 > 10 users - done, no errors, maps top load up to 49 > 15 users - done, no errors, maps top load up to 79 > 20 users - few timed out errors - connection timed out - maps top load > up to 78 > 25 users - more connection timed out errors - maps top load up to 73 > > no errors in geoserver.log > > non tiled > 1 user - done, no errors, maps top load up to 1 > 5 users - done, no errors - maps top load up to 4 - got kind of slow > 10 users - done, no errors , maps top load up to 9 - got kind of slow > > 12:46:36 up 35 days, 6:17, 5 users, load average: 9.66, 9.28, 8.39 > 492 processes: 481 sleeping, 11 running, 0 zombie, 0 stopped > CPU states: cpu user nice system irq softirq iowait idle > total 89.2% 0.0% 0.5% 0.0% 0.0% 0.7% 9.2% > cpu00 90.1% 0.0% 0.9% 0.0% 0.1% 1.5% 7.1% > cpu01 88.3% 0.0% 0.1% 0.0% 0.0% 0.0% 11.4% > Mem: 3034532k av, 2973636k used, 60896k free, 0k shrd, 122020k > buff > 2317924k actv, 404936k in_d, 48932k in_c > Swap: 9502400k av, 507784k used, 8994616k free 569560k > cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND > 3223 afreeman 20 0 1558M 1.5G 10588 R 20.8 51.4 1:41 1 java > 2949 afreeman 24 0 1558M 1.5G 10588 R 11.8 51.4 2:44 1 java > 2876 afreeman 20 0 1558M 1.5G 10588 R 10.7 51.4 2:38 1 java > 3301 afreeman 16 0 1558M 1.5G 10588 R 9.8 51.4 3:05 0 java > 1758 afreeman 15 0 1558M 1.5G 10588 S 9.6 51.4 1:55 0 java > 3224 afreeman 24 0 1558M 1.5G 10588 R 5.5 51.4 4:13 0 java > 3266 afreeman 24 0 1558M 1.5G 10588 R 5.4 51.4 2:01 0 java > 2292 afreeman 16 0 1558M 1.5G 10588 R 4.9 51.4 3:34 1 java > 2846 afreeman 25 0 1558M 1.5G 10588 S 4.5 51.4 3:16 0 java > 2797 afreeman 25 0 1558M 1.5G 10588 S 1.9 51.4 3:50 1 java > 1757 afreeman 16 0 1558M 1.5G 10588 S 1.1 51.4 1:09 1 java > 1756 afreeman 16 0 1558M 1.5G 10588 S 0.9 51.4 1:11 1 java > 1858 afreeman 15 0 1558M 1.5G 10588 S 0.4 51.4 0:05 0 java > 1909 afreeman 16 0 2332 2260 804 R 0.3 0.0 1:01 0 top > 1947 afreeman 15 0 1558M 1.5G 10588 S 0.2 51.4 0:06 0 java > 32063 jbosstom 15 0 257M 150M 9144 S 0.1 5.0 0:10 1 java > 1854 afreeman 15 0 1558M 1.5G 10588 S 0.1 51.4 0:05 0 java > 1 root 15 0 352 304 268 S 0.0 0.0 0:15 0 init > 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 > migration/0 > 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 > migration/1 > 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd > 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 > ksoftirqd/0 > 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 > ksoftirqd/1 > > > > > -- > Archive: http://lists.opengeo.org/[…]/1259613560921 > <http://lists.opengeo.org/massgis/archive/2009/11/1259613560921> > To unsubscribe send an email with subject "unsubscribe" to > massgis@... <mailto:massgis@...>. Please > contact massgis-manager@... > <mailto:massgis-manager@...> for questions. -- Gabriel Roldan OpenGeo - http://opengeo.org Expert service straight from the developers.
-
From: groldan
Subject: Re: More tests
Date: Dec 01, 2009 10:55 PM
Hi Aleda, please try out the attached jars. I did some changes that from the latest we tried, triples performance on arcy, and also improves predictability (like in the gap between min and max times being smaller). Please also make sure your JAI settings are correct in geoserver. This is how things are on arcy: Tomcat 5.5.28, no changes in config -Xms256 -Xmx512 JAI mem: 0.5 JAI threshold: 0.75 JAI threads: 7 JAI priority: 5 JPEG native: true PNG native: true I've ran the TILED tests with the following results: users: 1, threads: 12, avg.: 112ms, errors: 0 users: 5, threads: 60, avg.: 276ms, errors: 0 users: 10, threads: 120, avg.: 2018ms, errors: 0 users: 15, threads: 180, avg.: 2621ms, errors: 0 users: 20, threads: 240, avg: 4471, errors: 0 users: 25, threads: 300, avg: 8643, errors: 0 Those are better than the ones I've been having since a while (don't ask me exactly when, I tried so many things that's hard to keep a record of). One thing to consider too is that we might be hitting the server too hard wrt the actual number of threads the CPU can handle. I could rant more about that but am too tired now and need some sleep. One thing's sure, the tests should not break, and as long as performance for 1 and 5 users is ok on arcy, I would be eager to try out on at least one of the production machines. Let me know how it goes. Cheers, Gabriel
text.html (text/html) 8.8 kB
gt-arcsde-2.5-SNAPSHOT.jar (application/octet-stream) 292.6 kB
To unsubscribe from this list send an email to the list with the subject "unsubscribe" and an empty body.
Forgot your password?
Address:
Managers:
jdeolive (creator)
List Type:
Public List. Anyone who confirms their email address is valid can post and receive messages.
Archive:
The entire message, including attachments.