RE: [JBoss-user] JBossMQ Perforamance
i shouldn't need to go thereespecially considering one of the boxes in my gigabit tests runs the OSX and i'm not about to install linux for ppcwhat a step backwards! Plus all my users run w2k. -Original Message- From: Alwyn Schoeman [mailto:alwyn@;smart.com.ph] Sent: Thursday, October 24, 2002 9:21 PM To: [EMAIL PROTECTED] Subject: Re: [JBoss-user] JBossMQ Perforamance If you are using linux on both boxes connected directly with crossover-cable you might actually use special kernel functionality for that specific setup. Have not used it myself, but it is there if you want to look... On Thu, Oct 24, 2002 at 10:00:09AM -0500, Luttrell, Peter wrote: > Is there a way to reduce latency between 2 boxes connected with a 4 foot > crossover cable? Seams to me that this should provide the highest throughput > possible > > So are you basically saying that ~1500 messages per second is the fastest > that jbossmq can do? > > Did you see my message about hanging on to the connection which can lead to > >22,000 messages per second? > > .peter > > -Original Message- > From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] > Sent: Thursday, October 24, 2002 2:18 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > And if your issue was simply because of network latency (if my memory > serves, latency is the same on both type of networks)? > > ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network > latency? > > Cheers, > > > Sacha > > > -Message d'origine- > > De : [EMAIL PROTECTED] > > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > > Luttrell > > Envoye : mardi, 22 octobre 2002 07:41 > > A : [EMAIL PROTECTED] > > Objet : [JBoss-user] JBossMQ Perforamance > > > > > > In order to ascertain if JBossMQ is capable of providing the throughput > > I need, i've constructed a couple of little apps to see what kind of > > performance i can get. Here's what i've found: > > > > 100 Megabit:1250-1350 messages per second > > Gigabit:1500-1600 messages per second > > > > I was hoping to see a bit better performance then this; especially a > > larger differential with Gigabit. 100Megabit only used about 10% of the > > network bandwidth (if you believe xp's network monitor) and gigabit > > only used at most 1.5%. > > > > Is this the best performance i can expect? > > What have others observed? > > > > Can anyone suggest where the bottleneck might be? > > Does anyone have any suggestions on what configs to tweak? > > > > All of my test code, and deployable ear is located at > > http://www.sharpuniverse.com/jboss/jms-performance > > There is a publisher and subscriber swingapps which are webstart > > deployed. There is also a publisher servlet. Deploy the ear and go to > > context: jms-performance-test for everything > > > > Here's a little more info on my tests: > > JBoss3.0.3 > > Protocal/ConnecitonFactory: OIL > > 100MegaBit tests: > > Network: several different networks all switched > > Server/Client boxes: Various differnt boxes runing w2k, > > wxp, osx and > > linux, various different jdks 1.4.0 - 1.4.1_01 > > Gigabit tests: > > Network: 2 boxes with crossover cable > > Server: MacOSX running jdk1.3.1 > > Client: WindowsXP running jdk1.4.1_01 > > > > .peter > > > > > > > > --- > > This sf.net emial is sponsored by: Influence the future of > > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > > http://www.sun.com/javavote > > ___ > > JBoss-user mailing list > > [EMAIL PROTECTED] > > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > > > --- > This sf.net email is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4729346;7592162;s?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > This transmission contains information solely for intended recipient and may > be privileged, confidential and/or otherwise protect from disc
RE: [JBoss-user] JBossMQ Perforamance >> Hiram
Hiram? -Original Message- From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] Sent: Thursday, October 24, 2002 10:11 AM To: [EMAIL PROTECTED] Subject: RE: [JBoss-user] JBossMQ Perforamance I don't exactly know what is happening under the cover in this case. What I am saying is that if you have a single connection that does this: 1) take a message 2) send it 3) when sent, start again to point 1) Then, independently of the bandwith, you will have a latency limit that cannot be changed, event by using a 1Tb/s link. As an analogy, imagine that what you are trying to achieve is to deliver postal letters to your central office and that: - you have only one guy that can run between the local site and the destination (i.e. one connection) - this guy only deliver one message at a time - it takes 1 hour to go to the destination and come back (i.e. latency) No matter if the car is huge or if the highway is very wide: you will only be able to transport 24 letters per day (if the guy never sleeps). To go over this number, you need to use other scheme such as streaming the messages without waiting for an ACK, etc. Hiram is your guy anyway ;) Cheers, Sacha > -Message d'origine- > De : [EMAIL PROTECTED] > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Luttrell, > Peter > Envoyé : jeudi, 24 octobre 2002 17:00 > À : '[EMAIL PROTECTED]' > Objet : RE: [JBoss-user] JBossMQ Perforamance > > > Is there a way to reduce latency between 2 boxes connected with a 4 foot > crossover cable? Seams to me that this should provide the highest > throughput > possible > > So are you basically saying that ~1500 messages per second is the fastest > that jbossmq can do? > > Did you see my message about hanging on to the connection which > can lead to > >22,000 messages per second? > > .peter > > -Original Message- > From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] > Sent: Thursday, October 24, 2002 2:18 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > And if your issue was simply because of network latency (if my memory > serves, latency is the same on both type of networks)? > > ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network > latency? > > Cheers, > > > Sacha > > > -Message d'origine- > > De : [EMAIL PROTECTED] > > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > > Luttrell > > Envoye : mardi, 22 octobre 2002 07:41 > > A : [EMAIL PROTECTED] > > Objet : [JBoss-user] JBossMQ Perforamance > > > > > > In order to ascertain if JBossMQ is capable of providing the throughput > > I need, i've constructed a couple of little apps to see what kind of > > performance i can get. Here's what i've found: > > > > 100 Megabit:1250-1350 messages per second > > Gigabit:1500-1600 messages per second > > > > I was hoping to see a bit better performance then this; especially a > > larger differential with Gigabit. 100Megabit only used about 10% of the > > network bandwidth (if you believe xp's network monitor) and gigabit > > only used at most 1.5%. > > > > Is this the best performance i can expect? > > What have others observed? > > > > Can anyone suggest where the bottleneck might be? > > Does anyone have any suggestions on what configs to tweak? > > > > All of my test code, and deployable ear is located at > > http://www.sharpuniverse.com/jboss/jms-performance > > There is a publisher and subscriber swingapps which are webstart > > deployed. There is also a publisher servlet. Deploy the ear and go to > > context: jms-performance-test for everything > > > > Here's a little more info on my tests: > > JBoss3.0.3 > > Protocal/ConnecitonFactory: OIL > > 100MegaBit tests: > > Network: several different networks all switched > > Server/Client boxes: Various differnt boxes runing w2k, > > wxp, osx and > > linux, various different jdks 1.4.0 - 1.4.1_01 > > Gigabit tests: > > Network: 2 boxes with crossover cable > > Server: MacOSX running jdk1.3.1 > > Client: WindowsXP running jdk1.4.1_01 > > > > .peter > > > > > > > > --- > > This sf.net emial is sponsored by: Influence the future of > > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > > http://www.sun.com/javavote >
Re: [JBoss-user] JBossMQ Perforamance
If you are using linux on both boxes connected directly with crossover-cable you might actually use special kernel functionality for that specific setup. Have not used it myself, but it is there if you want to look... On Thu, Oct 24, 2002 at 10:00:09AM -0500, Luttrell, Peter wrote: > Is there a way to reduce latency between 2 boxes connected with a 4 foot > crossover cable? Seams to me that this should provide the highest throughput > possible > > So are you basically saying that ~1500 messages per second is the fastest > that jbossmq can do? > > Did you see my message about hanging on to the connection which can lead to > >22,000 messages per second? > > .peter > > -Original Message- > From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] > Sent: Thursday, October 24, 2002 2:18 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > And if your issue was simply because of network latency (if my memory > serves, latency is the same on both type of networks)? > > ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network > latency? > > Cheers, > > > Sacha > > > -Message d'origine- > > De : [EMAIL PROTECTED] > > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > > Luttrell > > Envoye : mardi, 22 octobre 2002 07:41 > > A : [EMAIL PROTECTED] > > Objet : [JBoss-user] JBossMQ Perforamance > > > > > > In order to ascertain if JBossMQ is capable of providing the throughput > > I need, i've constructed a couple of little apps to see what kind of > > performance i can get. Here's what i've found: > > > > 100 Megabit:1250-1350 messages per second > > Gigabit:1500-1600 messages per second > > > > I was hoping to see a bit better performance then this; especially a > > larger differential with Gigabit. 100Megabit only used about 10% of the > > network bandwidth (if you believe xp's network monitor) and gigabit > > only used at most 1.5%. > > > > Is this the best performance i can expect? > > What have others observed? > > > > Can anyone suggest where the bottleneck might be? > > Does anyone have any suggestions on what configs to tweak? > > > > All of my test code, and deployable ear is located at > > http://www.sharpuniverse.com/jboss/jms-performance > > There is a publisher and subscriber swingapps which are webstart > > deployed. There is also a publisher servlet. Deploy the ear and go to > > context: jms-performance-test for everything > > > > Here's a little more info on my tests: > > JBoss3.0.3 > > Protocal/ConnecitonFactory: OIL > > 100MegaBit tests: > > Network: several different networks all switched > > Server/Client boxes: Various differnt boxes runing w2k, > > wxp, osx and > > linux, various different jdks 1.4.0 - 1.4.1_01 > > Gigabit tests: > > Network: 2 boxes with crossover cable > > Server: MacOSX running jdk1.3.1 > > Client: WindowsXP running jdk1.4.1_01 > > > > .peter > > > > > > > > --- > > This sf.net emial is sponsored by: Influence the future of > > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > > http://www.sun.com/javavote > > ___ > > JBoss-user mailing list > > [EMAIL PROTECTED] > > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > > > --- > This sf.net email is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4729346;7592162;s?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > This transmission contains information solely for intended recipient and may > be privileged, confidential and/or otherwise protect from disclosure. If > you are not the intended recipient, please contact the sender and delete all > copies of this transmission. This message and/or the materials contained > herein are not an offer to sell, or a solicitation of an offer to buy, any > securities or other instruments. The information has been obtained or > derived from sources believed by us to be reliable, but we do not rep
RE: [JBoss-user] JBossMQ Perforamance
Is there a way to reduce latency between 2 boxes connected with a 4 foot crossover cable? Seams to me that this should provide the highest throughput possible So are you basically saying that ~1500 messages per second is the fastest that jbossmq can do? Did you see my message about hanging on to the connection which can lead to >22,000 messages per second? .peter -Original Message- From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] Sent: Thursday, October 24, 2002 2:18 AM To: [EMAIL PROTECTED] Subject: RE: [JBoss-user] JBossMQ Perforamance And if your issue was simply because of network latency (if my memory serves, latency is the same on both type of networks)? ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network latency? Cheers, Sacha > -Message d'origine- > De : [EMAIL PROTECTED] > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > Luttrell > Envoye : mardi, 22 octobre 2002 07:41 > A : [EMAIL PROTECTED] > Objet : [JBoss-user] JBossMQ Perforamance > > > In order to ascertain if JBossMQ is capable of providing the throughput > I need, i've constructed a couple of little apps to see what kind of > performance i can get. Here's what i've found: > > 100 Megabit: 1250-1350 messages per second > Gigabit: 1500-1600 messages per second > > I was hoping to see a bit better performance then this; especially a > larger differential with Gigabit. 100Megabit only used about 10% of the > network bandwidth (if you believe xp's network monitor) and gigabit > only used at most 1.5%. > > Is this the best performance i can expect? > What have others observed? > > Can anyone suggest where the bottleneck might be? > Does anyone have any suggestions on what configs to tweak? > > All of my test code, and deployable ear is located at > http://www.sharpuniverse.com/jboss/jms-performance > There is a publisher and subscriber swingapps which are webstart > deployed. There is also a publisher servlet. Deploy the ear and go to > context: jms-performance-test for everything > > Here's a little more info on my tests: > JBoss3.0.3 > Protocal/ConnecitonFactory: OIL > 100MegaBit tests: > Network: several different networks all switched > Server/Client boxes: Various differnt boxes runing w2k, > wxp, osx and > linux, various different jdks 1.4.0 - 1.4.1_01 > Gigabit tests: > Network: 2 boxes with crossover cable > Server: MacOSX running jdk1.3.1 > Client: WindowsXP running jdk1.4.1_01 > > .peter > > > > --- > This sf.net emial is sponsored by: Influence the future of > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > --- This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4729346;7592162;s?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user This transmission contains information solely for intended recipient and may be privileged, confidential and/or otherwise protect from disclosure. If you are not the intended recipient, please contact the sender and delete all copies of this transmission. This message and/or the materials contained herein are not an offer to sell, or a solicitation of an offer to buy, any securities or other instruments. The information has been obtained or derived from sources believed by us to be reliable, but we do not represent that it is accurate or complete. Any opinions or estimates contained in this information constitute our judgment as of this date and are subject to change without notice. Any information you share with us will be used in the operation of our business, and we do not request and do not want any material, nonpublic information. Absent an express prior written agreement, we are not agreeing to treat any information confidentially and will use any and all information and reserve the right to publish or disclose any information you share with us. --- This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4729346;7592162;s?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] JBossMQ Perforamance
I don't exactly know what is happening under the cover in this case. What I am saying is that if you have a single connection that does this: 1) take a message 2) send it 3) when sent, start again to point 1) Then, independently of the bandwith, you will have a latency limit that cannot be changed, event by using a 1Tb/s link. As an analogy, imagine that what you are trying to achieve is to deliver postal letters to your central office and that: - you have only one guy that can run between the local site and the destination (i.e. one connection) - this guy only deliver one message at a time - it takes 1 hour to go to the destination and come back (i.e. latency) No matter if the car is huge or if the highway is very wide: you will only be able to transport 24 letters per day (if the guy never sleeps). To go over this number, you need to use other scheme such as streaming the messages without waiting for an ACK, etc. Hiram is your guy anyway ;) Cheers, Sacha > -Message d'origine- > De : [EMAIL PROTECTED] > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Luttrell, > Peter > Envoyé : jeudi, 24 octobre 2002 17:00 > À : '[EMAIL PROTECTED]' > Objet : RE: [JBoss-user] JBossMQ Perforamance > > > Is there a way to reduce latency between 2 boxes connected with a 4 foot > crossover cable? Seams to me that this should provide the highest > throughput > possible > > So are you basically saying that ~1500 messages per second is the fastest > that jbossmq can do? > > Did you see my message about hanging on to the connection which > can lead to > >22,000 messages per second? > > .peter > > -Original Message- > From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] > Sent: Thursday, October 24, 2002 2:18 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > And if your issue was simply because of network latency (if my memory > serves, latency is the same on both type of networks)? > > ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network > latency? > > Cheers, > > > Sacha > > > -Message d'origine- > > De : [EMAIL PROTECTED] > > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > > Luttrell > > Envoye : mardi, 22 octobre 2002 07:41 > > A : [EMAIL PROTECTED] > > Objet : [JBoss-user] JBossMQ Perforamance > > > > > > In order to ascertain if JBossMQ is capable of providing the throughput > > I need, i've constructed a couple of little apps to see what kind of > > performance i can get. Here's what i've found: > > > > 100 Megabit:1250-1350 messages per second > > Gigabit:1500-1600 messages per second > > > > I was hoping to see a bit better performance then this; especially a > > larger differential with Gigabit. 100Megabit only used about 10% of the > > network bandwidth (if you believe xp's network monitor) and gigabit > > only used at most 1.5%. > > > > Is this the best performance i can expect? > > What have others observed? > > > > Can anyone suggest where the bottleneck might be? > > Does anyone have any suggestions on what configs to tweak? > > > > All of my test code, and deployable ear is located at > > http://www.sharpuniverse.com/jboss/jms-performance > > There is a publisher and subscriber swingapps which are webstart > > deployed. There is also a publisher servlet. Deploy the ear and go to > > context: jms-performance-test for everything > > > > Here's a little more info on my tests: > > JBoss3.0.3 > > Protocal/ConnecitonFactory: OIL > > 100MegaBit tests: > > Network: several different networks all switched > > Server/Client boxes: Various differnt boxes runing w2k, > > wxp, osx and > > linux, various different jdks 1.4.0 - 1.4.1_01 > > Gigabit tests: > > Network: 2 boxes with crossover cable > > Server: MacOSX running jdk1.3.1 > > Client: WindowsXP running jdk1.4.1_01 > > > > .peter > > > > > > > > --- > > This sf.net emial is sponsored by: Influence the future of > > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > > http://www.sun.com/javavote > > ___ > > JBoss-user mailing list > > [EMAIL PROTECTED] > > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > > >
RE: [JBoss-user] JBossMQ Perforamance
I understand what you're saying. Threading should solve this problem. Send 10 cars down your 10 car highway. I guess I would have thought that JbossMQ would already be threaded. Am I wrong? .peter -Original Message- From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] Sent: Thursday, October 24, 2002 10:11 AM To: [EMAIL PROTECTED] Subject: RE: [JBoss-user] JBossMQ Perforamance I don't exactly know what is happening under the cover in this case. What I am saying is that if you have a single connection that does this: 1) take a message 2) send it 3) when sent, start again to point 1) Then, independently of the bandwith, you will have a latency limit that cannot be changed, event by using a 1Tb/s link. As an analogy, imagine that what you are trying to achieve is to deliver postal letters to your central office and that: - you have only one guy that can run between the local site and the destination (i.e. one connection) - this guy only deliver one message at a time - it takes 1 hour to go to the destination and come back (i.e. latency) No matter if the car is huge or if the highway is very wide: you will only be able to transport 24 letters per day (if the guy never sleeps). To go over this number, you need to use other scheme such as streaming the messages without waiting for an ACK, etc. Hiram is your guy anyway ;) Cheers, Sacha > -Message d'origine- > De : [EMAIL PROTECTED] > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Luttrell, > Peter > Envoyé : jeudi, 24 octobre 2002 17:00 > À : '[EMAIL PROTECTED]' > Objet : RE: [JBoss-user] JBossMQ Perforamance > > > Is there a way to reduce latency between 2 boxes connected with a 4 foot > crossover cable? Seams to me that this should provide the highest > throughput > possible > > So are you basically saying that ~1500 messages per second is the fastest > that jbossmq can do? > > Did you see my message about hanging on to the connection which > can lead to > >22,000 messages per second? > > .peter > > -Original Message- > From: Sacha Labourey [mailto:Sacha.Labourey@;ml.cogito-info.ch] > Sent: Thursday, October 24, 2002 2:18 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > And if your issue was simply because of network latency (if my memory > serves, latency is the same on both type of networks)? > > ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network > latency? > > Cheers, > > > Sacha > > > -Message d'origine- > > De : [EMAIL PROTECTED] > > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > > Luttrell > > Envoye : mardi, 22 octobre 2002 07:41 > > A : [EMAIL PROTECTED] > > Objet : [JBoss-user] JBossMQ Perforamance > > > > > > In order to ascertain if JBossMQ is capable of providing the throughput > > I need, i've constructed a couple of little apps to see what kind of > > performance i can get. Here's what i've found: > > > > 100 Megabit:1250-1350 messages per second > > Gigabit:1500-1600 messages per second > > > > I was hoping to see a bit better performance then this; especially a > > larger differential with Gigabit. 100Megabit only used about 10% of the > > network bandwidth (if you believe xp's network monitor) and gigabit > > only used at most 1.5%. > > > > Is this the best performance i can expect? > > What have others observed? > > > > Can anyone suggest where the bottleneck might be? > > Does anyone have any suggestions on what configs to tweak? > > > > All of my test code, and deployable ear is located at > > http://www.sharpuniverse.com/jboss/jms-performance > > There is a publisher and subscriber swingapps which are webstart > > deployed. There is also a publisher servlet. Deploy the ear and go to > > context: jms-performance-test for everything > > > > Here's a little more info on my tests: > > JBoss3.0.3 > > Protocal/ConnecitonFactory: OIL > > 100MegaBit tests: > > Network: several different networks all switched > > Server/Client boxes: Various differnt boxes runing w2k, > > wxp, osx and > > linux, various different jdks 1.4.0 - 1.4.1_01 > > Gigabit tests: > > Network: 2 boxes with crossover cable > > Server: MacOSX running jdk1.3.1 > > Client: WindowsXP running jdk1.4.1_01 > > > > .peter > > > > > > > > --- > > This sf.net emial is sponsored by: Influence the future of
RE: [JBoss-user] JBossMQ Perforamance
And if your issue was simply because of network latency (if my memory serves, latency is the same on both type of networks)? ~1500 messages/sec => 1 message each 0.5/1ms. What is your actual network latency? Cheers, Sacha > -Message d'origine- > De : [EMAIL PROTECTED] > [mailto:jboss-user-admin@;lists.sourceforge.net]De la part de Peter > Luttrell > Envoye : mardi, 22 octobre 2002 07:41 > A : [EMAIL PROTECTED] > Objet : [JBoss-user] JBossMQ Perforamance > > > In order to ascertain if JBossMQ is capable of providing the throughput > I need, i've constructed a couple of little apps to see what kind of > performance i can get. Here's what i've found: > > 100 Megabit: 1250-1350 messages per second > Gigabit: 1500-1600 messages per second > > I was hoping to see a bit better performance then this; especially a > larger differential with Gigabit. 100Megabit only used about 10% of the > network bandwidth (if you believe xp's network monitor) and gigabit > only used at most 1.5%. > > Is this the best performance i can expect? > What have others observed? > > Can anyone suggest where the bottleneck might be? > Does anyone have any suggestions on what configs to tweak? > > All of my test code, and deployable ear is located at > http://www.sharpuniverse.com/jboss/jms-performance > There is a publisher and subscriber swingapps which are webstart > deployed. There is also a publisher servlet. Deploy the ear and go to > context: jms-performance-test for everything > > Here's a little more info on my tests: > JBoss3.0.3 > Protocal/ConnecitonFactory: OIL > 100MegaBit tests: > Network: several different networks all switched > Server/Client boxes: Various differnt boxes runing w2k, > wxp, osx and > linux, various different jdks 1.4.0 - 1.4.1_01 > Gigabit tests: > Network: 2 boxes with crossover cable > Server: MacOSX running jdk1.3.1 > Client: WindowsXP running jdk1.4.1_01 > > .peter > > > > --- > This sf.net emial is sponsored by: Influence the future of > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > --- This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4729346;7592162;s?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] JBossMQ Perforamance
that would explain the network utilization as all messages i'm working with are really small. with a very similar size 'message' i tried 2 additional tests with raw sockets. it's been a while since i've done work with sockets directly...so it's quite possible the much could be done to improve this test...also these little classes are single threaded and synchronous. attached is the code. here are the results: creating a new socket for each message: <850 messages per second reusing the same socket: >22,000 messages per second Is OIL holding on to the connections? If not, why not? .peter -Original Message- From: Rupp,Heiko [mailto:heiko.rupp@;bancotec.de] Sent: Tuesday, October 22, 2002 8:30 AM To: '[EMAIL PROTECTED]' Subject: RE: [JBoss-user] JBossMQ Perforamance > You should however have some difficulty filling up gigabit > ethernet from Packets per second is one thing. The other is packet size. With small packets, a network adapter or switch can be at its limit even if the network if not filled. On the other hand with large packets a network can fill up even if the throughput in pps is not at its maximum point. --- This sf.net emial is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user This transmission contains information solely for intended recipient and may be privileged, confidential and/or otherwise protect from disclosure. If you are not the intended recipient, please contact the sender and delete all copies of this transmission. This message and/or the materials contained herein are not an offer to sell, or a solicitation of an offer to buy, any securities or other instruments. The information has been obtained or derived from sources believed by us to be reliable, but we do not represent that it is accurate or complete. Any opinions or estimates contained in this information constitute our judgment as of this date and are subject to change without notice. Any information you share with us will be used in the operation of our business, and we do not request and do not want any material, nonpublic information. Absent an express prior written agreement, we are not agreeing to treat any information confidentially and will use any and all information and reserve the right to publish or disclose any information you share with us. socket-test.zip Description: Binary data
RE: [JBoss-user] JBossMQ Perforamance
Yes, i directly changed the class and hense tied it to the original OILs jndi name. Course in my case its ok as the only thing that is deployed on the instance is my performance test apps. But i quickly changed it back... -Original Message- From: Alwyn Schoeman [mailto:alwyn@;smart.com.ph] Sent: Tuesday, October 22, 2002 7:36 PM To: '[EMAIL PROTECTED]' Subject: Re: [JBoss-user] JBossMQ Perforamance Will this change mean that any programs will still continue to use "ConnectionFactory" in the JNDI lookup, but will get new OIL2 factory? >From which version is OIL2 supported? On Tue, Oct 22, 2002 at 11:26:54AM -0500, Luttrell, Peter wrote: > > I assume that all I need to change is from this: > > name="jboss.mq:service=InvocationLayer,type=OIL"> > optional-attribute-name="Invoker">jboss.mq:service=Invoker > ConnectionFactory > name="XAConnectionFactoryJNDIRef">XAConnectionFactory > 8090 > 6 > true > > > > to this: > > name="jboss.mq:service=InvocationLayer,type=OIL"> > optional-attribute-name="Invoker">jboss.mq:service=Invoker > ConnectionFactory > name="XAConnectionFactoryJNDIRef">XAConnectionFactory > 8090 > 6 > true > > > > If this is correct the results aren't so great, in fact its slower then the > original OIL. > > Just did a couple of tests at work (somewhat of a congested network): > > OIL1: 1000-1050 messages per second > OIL2: 550-600 messages per second > > .peter > > > > -Original Message- > From: Corby Page [mailto:CorbyPage@;duke-energy.com] > Sent: Tuesday, October 22, 2002 10:22 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > Peter, > > The new OIL2 Invocation layer is supposed to contain significant > performance enhancements. Plug in org.jboss.mq.il.oil2.OIL2ServerILService > as your new Invocation Layer and let us know the new results. > > Thanks, > Corby > > > > --- > This sf.net emial is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > This transmission contains information solely for intended recipient and may > be privileged, confidential and/or otherwise protect from disclosure. If > you are not the intended recipient, please contact the sender and delete all > copies of this transmission. This message and/or the materials contained > herein are not an offer to sell, or a solicitation of an offer to buy, any > securities or other instruments. The information has been obtained or > derived from sources believed by us to be reliable, but we do not represent > that it is accurate or complete. Any opinions or estimates contained in > this information constitute our judgment as of this date and are subject to > change without notice. Any information you share with us will be used in > the operation of our business, and we do not request and do not want any > material, nonpublic information. Absent an express prior written agreement, > we are not agreeing to treat any information confidentially and will use any > and all information and reserve the right to publish or disclose any > information you share with us. > > > --- > This sf.net emial is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." This transmission contains information solely for intended recipient and may be privileged, confidential and/or otherwise protect from disclosure. If you are not the intended recipient, please contact the sender and delete all copies of this transmission. This message and/or the materials contained herein are not an offer to sell, or a solicitation of an offer to buy, any securities or other instruments. The information has been obtained
Re: [JBoss-user] JBossMQ Perforamance
Will this change mean that any programs will still continue to use "ConnectionFactory" in the JNDI lookup, but will get new OIL2 factory? From which version is OIL2 supported? On Tue, Oct 22, 2002 at 11:26:54AM -0500, Luttrell, Peter wrote: > > I assume that all I need to change is from this: > > name="jboss.mq:service=InvocationLayer,type=OIL"> > optional-attribute-name="Invoker">jboss.mq:service=Invoker > ConnectionFactory > name="XAConnectionFactoryJNDIRef">XAConnectionFactory > 8090 > 6 > true > > > > to this: > > name="jboss.mq:service=InvocationLayer,type=OIL"> > optional-attribute-name="Invoker">jboss.mq:service=Invoker > ConnectionFactory > name="XAConnectionFactoryJNDIRef">XAConnectionFactory > 8090 > 6 > true > > > > If this is correct the results aren't so great, in fact its slower then the > original OIL. > > Just did a couple of tests at work (somewhat of a congested network): > > OIL1: 1000-1050 messages per second > OIL2: 550-600 messages per second > > .peter > > > > -Original Message- > From: Corby Page [mailto:CorbyPage@;duke-energy.com] > Sent: Tuesday, October 22, 2002 10:22 AM > To: [EMAIL PROTECTED] > Subject: RE: [JBoss-user] JBossMQ Perforamance > > > Peter, > > The new OIL2 Invocation layer is supposed to contain significant > performance enhancements. Plug in org.jboss.mq.il.oil2.OIL2ServerILService > as your new Invocation Layer and let us know the new results. > > Thanks, > Corby > > > > --- > This sf.net emial is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user > > > > This transmission contains information solely for intended recipient and may > be privileged, confidential and/or otherwise protect from disclosure. If > you are not the intended recipient, please contact the sender and delete all > copies of this transmission. This message and/or the materials contained > herein are not an offer to sell, or a solicitation of an offer to buy, any > securities or other instruments. The information has been obtained or > derived from sources believed by us to be reliable, but we do not represent > that it is accurate or complete. Any opinions or estimates contained in > this information constitute our judgment as of this date and are subject to > change without notice. Any information you share with us will be used in > the operation of our business, and we do not request and do not want any > material, nonpublic information. Absent an express prior written agreement, > we are not agreeing to treat any information confidentially and will use any > and all information and reserve the right to publish or disclose any > information you share with us. > > > --- > This sf.net emial is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22393/pgp0.pgp Description: PGP signature
RE: [JBoss-user] JBossMQ Perforamance
I assume that all I need to change is from this: jboss.mq:service=Invoker ConnectionFactory XAConnectionFactory 8090 6 true to this: jboss.mq:service=Invoker ConnectionFactory XAConnectionFactory 8090 6 true If this is correct the results aren't so great, in fact its slower then the original OIL. Just did a couple of tests at work (somewhat of a congested network): OIL1: 1000-1050 messages per second OIL2: 550-600 messages per second .peter -Original Message- From: Corby Page [mailto:CorbyPage@;duke-energy.com] Sent: Tuesday, October 22, 2002 10:22 AM To: [EMAIL PROTECTED] Subject: RE: [JBoss-user] JBossMQ Perforamance Peter, The new OIL2 Invocation layer is supposed to contain significant performance enhancements. Plug in org.jboss.mq.il.oil2.OIL2ServerILService as your new Invocation Layer and let us know the new results. Thanks, Corby --- This sf.net emial is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user This transmission contains information solely for intended recipient and may be privileged, confidential and/or otherwise protect from disclosure. If you are not the intended recipient, please contact the sender and delete all copies of this transmission. This message and/or the materials contained herein are not an offer to sell, or a solicitation of an offer to buy, any securities or other instruments. The information has been obtained or derived from sources believed by us to be reliable, but we do not represent that it is accurate or complete. Any opinions or estimates contained in this information constitute our judgment as of this date and are subject to change without notice. Any information you share with us will be used in the operation of our business, and we do not request and do not want any material, nonpublic information. Absent an express prior written agreement, we are not agreeing to treat any information confidentially and will use any and all information and reserve the right to publish or disclose any information you share with us. --- This sf.net emial is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] JBossMQ Perforamance
Peter, The new OIL2 Invocation layer is supposed to contain significant performance enhancements. Plug in org.jboss.mq.il.oil2.OIL2ServerILService as your new Invocation Layer and let us know the new results. Thanks, Corby --- This sf.net emial is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] JBossMQ Perforamance
> You should however have some difficulty filling up gigabit > ethernet from Packets per second is one thing. The other is packet size. With small packets, a network adapter or switch can be at its limit even if the network if not filled. On the other hand with large packets a network can fill up even if the throughput in pps is not at its maximum point. --- This sf.net emial is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ad.doubleclick.net/clk;4699841;7576301;v?http://www.sun.com/javavote ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] JBossMQ Perforamance
I must have been smoking. Seeing that the time slice per message will be shorter for Gigabit ethernet, the collision theory is mostly incorrect unless you have lots of users on same segment. You should however have some difficulty filling up gigabit ethernet from one pc. On Tue, Oct 22, 2002 at 12:40:42AM -0500, Peter Luttrell wrote: > In order to ascertain if JBossMQ is capable of providing the throughput > I need, i've constructed a couple of little apps to see what kind of > performance i can get. Here's what i've found: > > 100 Megabit: 1250-1350 messages per second > Gigabit: 1500-1600 messages per second > > I was hoping to see a bit better performance then this; especially a > larger differential with Gigabit. 100Megabit only used about 10% of the > network bandwidth (if you believe xp's network monitor) and gigabit > only used at most 1.5%. > > Is this the best performance i can expect? > What have others observed? > > Can anyone suggest where the bottleneck might be? > Does anyone have any suggestions on what configs to tweak? > > All of my test code, and deployable ear is located at > http://www.sharpuniverse.com/jboss/jms-performance > There is a publisher and subscriber swingapps which are webstart > deployed. There is also a publisher servlet. Deploy the ear and go to > context: jms-performance-test for everything > > Here's a little more info on my tests: > JBoss3.0.3 > Protocal/ConnecitonFactory: OIL > 100MegaBit tests: > Network: several different networks all switched > Server/Client boxes: Various differnt boxes runing w2k, wxp, osx and > linux, various different jdks 1.4.0 - 1.4.1_01 > Gigabit tests: > Network: 2 boxes with crossover cable > Server: MacOSX running jdk1.3.1 > Client: WindowsXP running jdk1.4.1_01 > > .peter > > > > --- > This sf.net emial is sponsored by: Influence the future of > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22375/pgp0.pgp Description: PGP signature
Re: [JBoss-user] JBossMQ Perforamance
Have you tried multiple client connections? I'm also thinking that maybe the probability of collisions on Gigabit ethernet is the same as on 100Mbit as it is not governed by capacity but time. It could actually mean that it is not possible to send that much more packets on Gigabit than on 100Mbit. This is then of course influenced by the collision domains of your network. Also have you taken into account the actual io and cpu usage on the machines? If the load or io on the machine gets higher it will limit the amount of messages that you can put on a queue. On Tue, Oct 22, 2002 at 12:40:42AM -0500, Peter Luttrell wrote: > In order to ascertain if JBossMQ is capable of providing the throughput > I need, i've constructed a couple of little apps to see what kind of > performance i can get. Here's what i've found: > > 100 Megabit: 1250-1350 messages per second > Gigabit: 1500-1600 messages per second > > I was hoping to see a bit better performance then this; especially a > larger differential with Gigabit. 100Megabit only used about 10% of the > network bandwidth (if you believe xp's network monitor) and gigabit > only used at most 1.5%. > > Is this the best performance i can expect? > What have others observed? > > Can anyone suggest where the bottleneck might be? > Does anyone have any suggestions on what configs to tweak? > > All of my test code, and deployable ear is located at > http://www.sharpuniverse.com/jboss/jms-performance > There is a publisher and subscriber swingapps which are webstart > deployed. There is also a publisher servlet. Deploy the ear and go to > context: jms-performance-test for everything > > Here's a little more info on my tests: > JBoss3.0.3 > Protocal/ConnecitonFactory: OIL > 100MegaBit tests: > Network: several different networks all switched > Server/Client boxes: Various differnt boxes runing w2k, wxp, osx and > linux, various different jdks 1.4.0 - 1.4.1_01 > Gigabit tests: > Network: 2 boxes with crossover cable > Server: MacOSX running jdk1.3.1 > Client: WindowsXP running jdk1.4.1_01 > > .peter > > > > --- > This sf.net emial is sponsored by: Influence the future of > Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) > program now. http://ad.doubleclick.net/clk;4699841;7576301;v? > http://www.sun.com/javavote > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22373/pgp0.pgp Description: PGP signature