RE: [JBoss-user] httpsession clustering overhead
ECPERF is in CVS under the ecperf module. For testing with 3.0 series, you'll need to checkout: ecperf -r Branch_3_0 3.2 and higher, just check out ecperf regulary. at jboss's CVS server? sorry for asking, but me cvs aren't that close so that I fully understand all you said ;-) Also ECPerf is not a good HTTP test. For that I suggest the RICE study (JBoss vs. Jonas). Dain has done some great work porting the test to 3.0, but we haven't committed it yet. Any link to it? At a first glance I found the rice university page, but not the study. Any ideia when the porting will be commited? Could anybody offer up machines for doing clustering benchmarks with RICE and ECPERF? I only have 3 machines at home and its just not enough. Email me if you're interested. How many are needed? At my research place I have 4 machines that I can use however I want, with Win2k or Linux. Besides that I have some access to some servers and about other 5 machines running Win2K. I would really like to do those benchmarck since I will use them in my thesis.. Thank you Joao Clemente Inesc-Id - Portugal --- This SF.net email is sponsored by: ApacheCon, November 18-21 in Las Vegas (supported by COMDEX), the only Apache event to be fully supported by the ASF. http://www.apachecon.com ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] httpsession clustering overhead
Joao Pedro Clemente wrote: ECPERF is in CVS under the ecperf module. For testing with 3.0 series, you'll need to checkout: ecperf -r Branch_3_0 3.2 and higher, just check out ecperf regulary. at jboss's CVS server? sorry for asking, but me cvs aren't that close so that I fully understand all you said ;-) Yes, it's in the jboss cvs server. Also ECPerf is not a good HTTP test. For that I suggest the RICE study (JBoss vs. Jonas). Dain has done some great work porting the test to 3.0, but we haven't committed it yet. Any link to it? At a first glance I found the rice university page, but not the study. Any ideia when the porting will be commited? The porting is done, but it is no to easy to setup and run. I doubt it will be committed any time soon. -dain --- This SF.net email is sponsored by: ApacheCon, November 18-21 in Las Vegas (supported by COMDEX), the only Apache event to be fully supported by the ASF. http://www.apachecon.com ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] httpsession clustering overhead
I am beginning to get httpclustering to work. Aside from the overhead of each server having to have the capacity to hold the total sessions in all servers, what's the wire traffic like? Yea, I'll sniff it if I have to, this is just to see if anyone has a quick idea. Thats a good point that I once was worried about also. I even tought of setting up a network infrastructure with a backplane only for those messages. In the messages I have exchenged in the forum ML, I got the feeling that for usual setup a 100 Mbps network is more than enough to handle the replication messages without a problem. Netvertheless, I have read lots of J2EE design recomendation (generic, not particular to JBoss) that would recomend to design your structures so that you would need to sync the smallest possible info. I am sorry I cannot answer your question right now, as I still haven't found a way to benchmark the efficiency of a single Jboss 3.x, in a way to find the improvment that a clustered environment can get. BTW, has someone got ECPerf to work with JBoss 3.x? I have read a posting from bill burk were I got the feeling he has done it, but every time I ask about this everyone points me to that project that has the instructions for JBoss 2.4.x series.. Joao Clemente --- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] httpsession clustering overhead
I did think of the backplane idea as well. But we also have memory consumption to think about. If I had four servers evenly balanced, say 100 users on each, I think I would end up with 400 sessions per server. Our sessions are currently pigs, and that is motivation to fix that! Perhaps better would be the cmp implementation which I haven't tried. There, it seems the servers could keep only the load they need, and if the get an unknown session, the cmp would pull in the data if needed. I'm not sure it really works that way though, I suppose it depends on how bean clustering works, and I haven't looked into that. It could be just as bad if bean clustering works in high availability mode as well! Thats a good point that I once was worried about also. I even tought of setting up a network infrastructure with a backplane only for those messages. In the messages I have exchenged in the forum ML, I got the feeling that for usual setup a 100 Mbps network is more than enough to handle the replication messages without a problem. Netvertheless, I have read lots of J2EE design recomendation (generic, not particular to JBoss) that would recomend to design your structures so that you would need to sync the smallest possible info. I am sorry I cannot answer your question right now, as I still haven't found a way to benchmark the efficiency of a single Jboss 3.x, in a way to find the improvment that a clustered environment can get. --- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] httpsession clustering overhead
ECPERF is in CVS under the ecperf module. For testing with 3.0 series, you'll need to checkout: ecperf -r Branch_3_0 3.2 and higher, just check out ecperf regulary. It is configured to use either BMP, CMP, or clustering. You can email me for help, but I'm pretty busy right now so I may not respond to lengthy questions. Also ECPerf is not a good HTTP test. For that I suggest the RICE study (JBoss vs. Jonas). Dain has done some great work porting the test to 3.0, but we haven't committed it yet. Could anybody offer up machines for doing clustering benchmarks with RICE and ECPERF? I only have 3 machines at home and its just not enough. Email me if you're interested. Bill -Original Message- From: [EMAIL PROTECTED] [mailto:jboss-user-admin;lists.sourceforge.net]On Behalf Of Rick LaBanca Sent: Saturday, November 02, 2002 12:31 PM To: [EMAIL PROTECTED] Subject: Re: [JBoss-user] httpsession clustering overhead I did think of the backplane idea as well. But we also have memory consumption to think about. If I had four servers evenly balanced, say 100 users on each, I think I would end up with 400 sessions per server. Our sessions are currently pigs, and that is motivation to fix that! Perhaps better would be the cmp implementation which I haven't tried. There, it seems the servers could keep only the load they need, and if the get an unknown session, the cmp would pull in the data if needed. I'm not sure it really works that way though, I suppose it depends on how bean clustering works, and I haven't looked into that. It could be just as bad if bean clustering works in high availability mode as well! Thats a good point that I once was worried about also. I even tought of setting up a network infrastructure with a backplane only for those messages. In the messages I have exchenged in the forum ML, I got the feeling that for usual setup a 100 Mbps network is more than enough to handle the replication messages without a problem. Netvertheless, I have read lots of J2EE design recomendation (generic, not particular to JBoss) that would recomend to design your structures so that you would need to sync the smallest possible info. I am sorry I cannot answer your question right now, as I still haven't found a way to benchmark the efficiency of a single Jboss 3.x, in a way to find the improvment that a clustered environment can get. --- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user --- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user