JVM tuning
Has anyone got any clues on JVM tuning for Tomcat on Windows 2000, Sun 141 or 142? Im playing with some of the young/old generation and GC settings but am having trouble deciding what is best. I think that I should probably push out the young generation to be a larger than normal amount as many of the objects created are very transitory, and I have seen lots of GC activity going on there. On that note, has anyone tried the Sun GC portal? Im a bit baffled as to how to create a log file for it - turning on the switches is all very well and prints to the screen OK but it doesnt seem to log to the .out files? cheers Pete
RE: JVM tuning
Howdy, >Has anyone got any clues on JVM tuning for Tomcat on Windows 2000, Sun 141 I would venture at least a few people on this list have clues. But it's impossible to help you without a clue about your webapp. Yoav Shapira This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Hmm well only a profiler could tell me that; I was more looking for some (subsequently discovered) things such as you should reseize the young generation to be much larger than the default 25% and so on; there are some rules of thumb such as this which I was looking for! cheers Pete "Shapira, Yoav" <[EMAIL PROTECTED]> 04/08/2003 14:16 Please respond to "Tomcat Users List" To: "Tomcat Users List" <[EMAIL PROTECTED]> cc: Subject:RE: JVM tuning Howdy, >Has anyone got any clues on JVM tuning for Tomcat on Windows 2000, Sun 141 I would venture at least a few people on this list have clues. But it's impossible to help you without a clue about your webapp. Yoav Shapira This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Howdy, >-Original Message- >From: [EMAIL PROTECTED] >[mailto:[EMAIL PROTECTED] >Sent: Monday, August 04, 2003 9:27 AM >To: Tomcat Users List >Subject: RE: JVM tuning > >Hmm well only a profiler could tell me that; I was more looking for some >(subsequently discovered) things such as you should reseize the young >generation to be much larger than the default 25% and so on; there are >some rules of thumb such as this which I was looking for! I would argue any such rule of thumb is as likely to hurt performance as improve it for your specific app. Yoav Shapira Millennium ChemInformatics This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Well not really; we know that we are running Tomcat, a web container which has its own (fixed) characteristics. It is a server side app which is processing non state based transactions which are thus highly like to involve a lot of objects being created and destroyed without too many hanging around for long; the details of the webapp, unless it is highly unusual, are likely not to matter particularly. This is a rule of thumb not an exact science and the vast majority of people who run Tomcat would benefit from running in such a configuration (or playing with it to see what the effects were) - read Sun's own tuning documentation and you will see the default settings are not said to be suitable for the majority of server apps. cheers Pete "Shapira, Yoav" <[EMAIL PROTECTED]> 04/08/2003 14:29 Please respond to "Tomcat Users List" To: "Tomcat Users List" <[EMAIL PROTECTED]> cc: Subject:RE: JVM tuning Howdy, >-Original Message- >From: [EMAIL PROTECTED] >[mailto:[EMAIL PROTECTED] >Sent: Monday, August 04, 2003 9:27 AM >To: Tomcat Users List >Subject: RE: JVM tuning > >Hmm well only a profiler could tell me that; I was more looking for some >(subsequently discovered) things such as you should reseize the young >generation to be much larger than the default 25% and so on; there are >some rules of thumb such as this which I was looking for! I would argue any such rule of thumb is as likely to hurt performance as improve it for your specific app. Yoav Shapira Millennium ChemInformatics This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Howdy, >Well not really; we know that we are running Tomcat, a web container which >has its own (fixed) characteristics. It is a server side app which is >processing non state based transactions which are thus highly like to >involve a lot of objects being created and destroyed without too many >hanging around for long; the details of the webapp, unless it is highly >unusual, are likely not to matter particularly. This is a rule of thumb >not an exact science and the vast majority of people who run Tomcat would >benefit from running in such a configuration (or playing with it to see >what the effects were) I disagree ;) Every time I've tuned a webapp for pay, both transactional and not, both full J2EE and "just" servlets/JSPs, the above has been false. The webapp's specific implementation matters far more (typically 3-4 orders of magnitude) than the container implementation, especially for a mature container which has been tuned repetitively and carefully over time. It's precisely because of this, and because as you say performance tuning is not a science, that a rule of thumb is as likely to hurt as it is to help if blindly applied. > read Sun's own tuning documentation and you will >see the default settings are not said to be suitable for the majority of >server apps. I appreciate the pointer - I used to help write them ;) Good luck, however, as it's always YMMV with these things. Yoav Shapira This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Ah sorry I should really have said that I was focusing specifically on memory managment; general performance tuning is a completely different matter in which I do agree with you! On the memory front however I would stand by the fact that most tomcat webapps (big or small) would benefit from a change in the default JVM config. Anyway, have sorted out my problem now (by doing just that as well as a GC change) and performance is 10x better! cheers Pete "Shapira, Yoav" <[EMAIL PROTECTED]> 04/08/2003 14:45 Please respond to "Tomcat Users List" To: "Tomcat Users List" <[EMAIL PROTECTED]> cc: Subject:RE: JVM tuning Howdy, >Well not really; we know that we are running Tomcat, a web container which >has its own (fixed) characteristics. It is a server side app which is >processing non state based transactions which are thus highly like to >involve a lot of objects being created and destroyed without too many >hanging around for long; the details of the webapp, unless it is highly >unusual, are likely not to matter particularly. This is a rule of thumb >not an exact science and the vast majority of people who run Tomcat would >benefit from running in such a configuration (or playing with it to see >what the effects were) I disagree ;) Every time I've tuned a webapp for pay, both transactional and not, both full J2EE and "just" servlets/JSPs, the above has been false. The webapp's specific implementation matters far more (typically 3-4 orders of magnitude) than the container implementation, especially for a mature container which has been tuned repetitively and carefully over time. It's precisely because of this, and because as you say performance tuning is not a science, that a rule of thumb is as likely to hurt as it is to help if blindly applied. > read Sun's own tuning documentation and you will >see the default settings are not said to be suitable for the majority of >server apps. I appreciate the pointer - I used to help write them ;) Good luck, however, as it's always YMMV with these things. Yoav Shapira This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Wouldn't AdaptiveSizePolicy help? (saves you the work of Java Heap usage analyzing :-) : I use this on my 2x proc machine. -XX:UseParallelGC -XX:+UseAdaptiveSizePolicy > -Original Message- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] > Sent: Monday, August 04, 2003 9:55 AM > To: Tomcat Users List > Subject: RE: JVM tuning > > > Ah sorry I should really have said that I was focusing > specifically on > memory managment; general performance tuning is a completely > different > matter in which I do agree with you! On the memory front > however I would > stand by the fact that most tomcat webapps (big or small) > would benefit > from a change in the default JVM config. > > Anyway, have sorted out my problem now (by doing just that as > well as a GC > change) and performance is 10x better! > > cheers > Pete > > > > > > "Shapira, Yoav" <[EMAIL PROTECTED]> > 04/08/2003 14:45 > Please respond to "Tomcat Users List" > > To: "Tomcat Users List" <[EMAIL PROTECTED]> > cc: > Subject:RE: JVM tuning > > > > Howdy, > > >Well not really; we know that we are running Tomcat, a web container > which > >has its own (fixed) characteristics. It is a server side > app which is > >processing non state based transactions which are thus highly like to > >involve a lot of objects being created and destroyed without too many > >hanging around for long; the details of the webapp, unless it is > highly > >unusual, are likely not to matter particularly. This is a rule of > thumb > >not an exact science and the vast majority of people who run Tomcat > would > >benefit from running in such a configuration (or playing > with it to see > >what the effects were) > > I disagree ;) > > Every time I've tuned a webapp for pay, both transactional > and not, both > full J2EE and "just" servlets/JSPs, the above has been false. The > webapp's specific implementation matters far more (typically > 3-4 orders > of magnitude) than the container implementation, especially > for a mature > container which has been tuned repetitively and carefully over time. > It's precisely because of this, and because as you say performance > tuning is not a science, that a rule of thumb is as likely to > hurt as it > is to help if blindly applied. > > > read Sun's own tuning documentation and you will > >see the default settings are not said to be suitable for the majority > of > >server apps. > > I appreciate the pointer - I used to help write them ;) > > Good luck, however, as it's always YMMV with these things. > > Yoav Shapira > > > > This e-mail, including any attachments, is a confidential business > communication, and may contain information that is confidential, > proprietary and/or privileged. This e-mail is intended only for the > individual(s) to whom it is addressed, and may not be saved, copied, > printed, disclosed or used by anyone else. If you are not the(an) > intended recipient, please immediately delete this e-mail from your > computer system and notify the sender. Thank you. > > > - > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Isnt the adaptive sizing only relevant to much larger memory configurations (Im running this JVM with a max heap of 1-1.5Gb? Pete "Angus Mezick" <[EMAIL PROTECTED]> 04/08/2003 17:51 Please respond to "Tomcat Users List" To: "Tomcat Users List" <[EMAIL PROTECTED]> cc: Subject:RE: JVM tuning Wouldn't AdaptiveSizePolicy help? (saves you the work of Java Heap usage analyzing :-) : I use this on my 2x proc machine. -XX:UseParallelGC -XX:+UseAdaptiveSizePolicy > -Original Message- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] > Sent: Monday, August 04, 2003 9:55 AM > To: Tomcat Users List > Subject: RE: JVM tuning > > > Ah sorry I should really have said that I was focusing > specifically on > memory managment; general performance tuning is a completely > different > matter in which I do agree with you! On the memory front > however I would > stand by the fact that most tomcat webapps (big or small) > would benefit > from a change in the default JVM config. > > Anyway, have sorted out my problem now (by doing just that as > well as a GC > change) and performance is 10x better! > > cheers > Pete > > > > > > "Shapira, Yoav" <[EMAIL PROTECTED]> > 04/08/2003 14:45 > Please respond to "Tomcat Users List" > > To: "Tomcat Users List" <[EMAIL PROTECTED]> > cc: > Subject:RE: JVM tuning > > > > Howdy, > > >Well not really; we know that we are running Tomcat, a web container > which > >has its own (fixed) characteristics. It is a server side > app which is > >processing non state based transactions which are thus highly like to > >involve a lot of objects being created and destroyed without too many > >hanging around for long; the details of the webapp, unless it is > highly > >unusual, are likely not to matter particularly. This is a rule of > thumb > >not an exact science and the vast majority of people who run Tomcat > would > >benefit from running in such a configuration (or playing > with it to see > >what the effects were) > > I disagree ;) > > Every time I've tuned a webapp for pay, both transactional > and not, both > full J2EE and "just" servlets/JSPs, the above has been false. The > webapp's specific implementation matters far more (typically > 3-4 orders > of magnitude) than the container implementation, especially > for a mature > container which has been tuned repetitively and carefully over time. > It's precisely because of this, and because as you say performance > tuning is not a science, that a rule of thumb is as likely to > hurt as it > is to help if blindly applied. > > > read Sun's own tuning documentation and you will > >see the default settings are not said to be suitable for the majority > of > >server apps. > > I appreciate the pointer - I used to help write them ;) > > Good luck, however, as it's always YMMV with these things. > > Yoav Shapira > > > > This e-mail, including any attachments, is a confidential business > communication, and may contain information that is confidential, > proprietary and/or privileged. This e-mail is intended only for the > individual(s) to whom it is addressed, and may not be saved, copied, > printed, disclosed or used by anyone else. If you are not the(an) > intended recipient, please immediately delete this e-mail from your > computer system and notify the sender. Thank you. > > > - > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: JVM tuning
Ok, so if I set to 512M it isn't that effective? Would this be better in a 2xproc 512M heap environment: ConcurrentGC with ParNewGC (ParNewGC on Multi-CPU machines): -XX:+UseConcMarkSweepGC -XX:+UseParNewGC > -Original Message- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] > Sent: Monday, August 04, 2003 1:03 PM > To: Tomcat Users List > Subject: RE: JVM tuning > > > Isnt the adaptive sizing only relevant to much larger memory > configurations (Im running this JVM with a max heap of 1-1.5Gb? > Pete > > > > > > "Angus Mezick" <[EMAIL PROTECTED]> > 04/08/2003 17:51 > Please respond to "Tomcat Users List" > > To: "Tomcat Users List" <[EMAIL PROTECTED]> > cc: > Subject:RE: JVM tuning > > > Wouldn't AdaptiveSizePolicy help? (saves you the work of Java > Heap usage > analyzing :-) : > > I use this on my 2x proc machine. > > -XX:UseParallelGC -XX:+UseAdaptiveSizePolicy > > > -Original Message- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] > > Sent: Monday, August 04, 2003 9:55 AM > > To: Tomcat Users List > > Subject: RE: JVM tuning > > > > > > Ah sorry I should really have said that I was focusing > > specifically on > > memory managment; general performance tuning is a completely > > different > > matter in which I do agree with you! On the memory front > > however I would > > stand by the fact that most tomcat webapps (big or small) > > would benefit > > from a change in the default JVM config. > > > > Anyway, have sorted out my problem now (by doing just that as > > well as a GC > > change) and performance is 10x better! > > > > cheers > > Pete > > > > > > > > > > > > "Shapira, Yoav" <[EMAIL PROTECTED]> > > 04/08/2003 14:45 > > Please respond to "Tomcat Users List" > > > > To: "Tomcat Users List" <[EMAIL PROTECTED]> > > cc: > > Subject:RE: JVM tuning > > > > > > > > Howdy, > > > > >Well not really; we know that we are running Tomcat, a web > container > > which > > >has its own (fixed) characteristics. It is a server side > > app which is > > >processing non state based transactions which are thus > highly like to > > >involve a lot of objects being created and destroyed > without too many > > >hanging around for long; the details of the webapp, unless it is > > highly > > >unusual, are likely not to matter particularly. This is a rule of > > thumb > > >not an exact science and the vast majority of people who run Tomcat > > would > > >benefit from running in such a configuration (or playing > > with it to see > > >what the effects were) > > > > I disagree ;) > > > > Every time I've tuned a webapp for pay, both transactional > > and not, both > > full J2EE and "just" servlets/JSPs, the above has been false. The > > webapp's specific implementation matters far more (typically > > 3-4 orders > > of magnitude) than the container implementation, especially > > for a mature > > container which has been tuned repetitively and carefully over time. > > It's precisely because of this, and because as you say performance > > tuning is not a science, that a rule of thumb is as likely to > > hurt as it > > is to help if blindly applied. > > > > > read Sun's own tuning documentation and you will > > >see the default settings are not said to be suitable for > the majority > > of > > >server apps. > > > > I appreciate the pointer - I used to help write them ;) > > > > Good luck, however, as it's always YMMV with these things. > > > > Yoav Shapira > > > > > > > > This e-mail, including any attachments, is a confidential business > > communication, and may contain information that is confidential, > > proprietary and/or privileged. This e-mail is intended > only for the > > individual(s) to whom it is addressed, and may not be > saved, copied, > > printed, disclosed or used by anyone else. If you are not the(an) > > intended recipient, please immediately delete this e-mail from your > > computer system and notify the sender. Thank you. > > > > > > > - > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > > > - > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]