Re: T4.1.5 possible ognl cache problem
No, i've checked all my html (it's not a single html because we're using lots of components). -- denis On the page in your production application, you wouldn't be using #this anywhere in an OGNL expression, would you? Kalle On Wed, Sep 3, 2008 at 6:35 AM, denis queffeulou < [EMAIL PROTECTED]> wrote: I've done a simple webapp and the result is that compileExpression is called 18 times but this doesn't increase. So for that webapp it is Ok. Unfortunately, my main app still have the problem (org.apache.tapestry.disable-caching is not set so takes its default value). The last test I used is to request a single page and the number of calls to compileExpression is the same as the request number (minus 1). It takes about 16% cpu (it depends on the complexity of the page). I don't know what is so different between the two webapps that the ognl cache (something not used directly by my code) seems not to be working. -- denis are you sure, you have org.apache.tapestry.disable-caching set to false? You can easily check this by looking at Tapestry's exception page. 2008/9/1 denis queffeulou <[EMAIL PROTECTED]> h, actually my WAR is a big application and I can't put it in Jira. Don't you already have pages with some OGNL expression in test case ? I look at the source code and see that a WeakHashMap but even with half memory used, the cache hits are not numerous. Is there any configuration to tweak ? (I'm not in debug mode). Perhaps it's not a bug. Anyway I've made a issue in Jira (2635). Can you attach a War-File to a an issue in Jira? If so, I promise to have a look at it :). 2008/9/1 denis queffeulou <[EMAIL PROTECTED]> I've made another test with 4.1.5 which gives: 34,2% 90528 invocations on ExpressionCache.getCompiledExpression 25,4% 90300 invocations on Ognl.compileExpression The test is made of 5 pages which returns XML and has run for about 2 hours. I'm not sure to be able to test with 4.1.6 in the next days. -- Denis Jesse, could you comment on this? I haven't tried profiling my apps, but the repeated calls to Ognl.compileExpression are worrisome if Denis got his numbers right. Are there any ognl-related changes in 4.1.6? Denis, maybe you could repeat your profiling with the latest snapshot, so hopefully something could still be done for this before 4.1.6 is released. Kalle On Thu, Aug 28, 2008 at 5:08 AM, denis queffeulou < [EMAIL PROTECTED]> wrote: Hi all, I'm making some profiling on an web app made with Tapestry 4.1.5 which returns XML. The result are disappointing as the average access time (got with JMeter) is around 2000ms. In JProfiler, I found that OGNL expressions compilation uses 40% of processing time. As the JMeter script is requesting always the same page on the webapp, I'm wondering why the expressions are always compiled. The cache seems not to be used: 31896 invocations on ExpressionCache.getCompiledExpression 31140 invocations on Ognl.compileExpression Is there someone to give me a clue to have better performances ? Thanks PS: sorry for double posting in dev list, I had some problems to subscribe to users. -- Denis Queffeulou - 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] - 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: T4.1.5 possible ognl cache problem
I've done a simple webapp and the result is that compileExpression is called 18 times but this doesn't increase. So for that webapp it is Ok. Unfortunately, my main app still have the problem (org.apache.tapestry.disable-caching is not set so takes its default value). The last test I used is to request a single page and the number of calls to compileExpression is the same as the request number (minus 1). It takes about 16% cpu (it depends on the complexity of the page). I don't know what is so different between the two webapps that the ognl cache (something not used directly by my code) seems not to be working. -- denis are you sure, you have org.apache.tapestry.disable-caching set to false? You can easily check this by looking at Tapestry's exception page. 2008/9/1 denis queffeulou <[EMAIL PROTECTED]> h, actually my WAR is a big application and I can't put it in Jira. Don't you already have pages with some OGNL expression in test case ? I look at the source code and see that a WeakHashMap but even with half memory used, the cache hits are not numerous. Is there any configuration to tweak ? (I'm not in debug mode). Perhaps it's not a bug. Anyway I've made a issue in Jira (2635). Can you attach a War-File to a an issue in Jira? If so, I promise to have a look at it :). 2008/9/1 denis queffeulou <[EMAIL PROTECTED]> I've made another test with 4.1.5 which gives: 34,2% 90528 invocations on ExpressionCache.getCompiledExpression 25,4% 90300 invocations on Ognl.compileExpression The test is made of 5 pages which returns XML and has run for about 2 hours. I'm not sure to be able to test with 4.1.6 in the next days. -- Denis Jesse, could you comment on this? I haven't tried profiling my apps, but the repeated calls to Ognl.compileExpression are worrisome if Denis got his numbers right. Are there any ognl-related changes in 4.1.6? Denis, maybe you could repeat your profiling with the latest snapshot, so hopefully something could still be done for this before 4.1.6 is released. Kalle On Thu, Aug 28, 2008 at 5:08 AM, denis queffeulou < [EMAIL PROTECTED]> wrote: Hi all, I'm making some profiling on an web app made with Tapestry 4.1.5 which returns XML. The result are disappointing as the average access time (got with JMeter) is around 2000ms. In JProfiler, I found that OGNL expressions compilation uses 40% of processing time. As the JMeter script is requesting always the same page on the webapp, I'm wondering why the expressions are always compiled. The cache seems not to be used: 31896 invocations on ExpressionCache.getCompiledExpression 31140 invocations on Ognl.compileExpression Is there someone to give me a clue to have better performances ? Thanks PS: sorry for double posting in dev list, I had some problems to subscribe to users. -- Denis Queffeulou - 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] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: T4.1.5 possible ognl cache problem
h, actually my WAR is a big application and I can't put it in Jira. Don't you already have pages with some OGNL expression in test case ? I look at the source code and see that a WeakHashMap but even with half memory used, the cache hits are not numerous. Is there any configuration to tweak ? (I'm not in debug mode). Perhaps it's not a bug. Anyway I've made a issue in Jira (2635). Can you attach a War-File to a an issue in Jira? If so, I promise to have a look at it :). 2008/9/1 denis queffeulou <[EMAIL PROTECTED]> I've made another test with 4.1.5 which gives: 34,2% 90528 invocations on ExpressionCache.getCompiledExpression 25,4% 90300 invocations on Ognl.compileExpression The test is made of 5 pages which returns XML and has run for about 2 hours. I'm not sure to be able to test with 4.1.6 in the next days. -- Denis Jesse, could you comment on this? I haven't tried profiling my apps, but the repeated calls to Ognl.compileExpression are worrisome if Denis got his numbers right. Are there any ognl-related changes in 4.1.6? Denis, maybe you could repeat your profiling with the latest snapshot, so hopefully something could still be done for this before 4.1.6 is released. Kalle On Thu, Aug 28, 2008 at 5:08 AM, denis queffeulou < [EMAIL PROTECTED]> wrote: Hi all, I'm making some profiling on an web app made with Tapestry 4.1.5 which returns XML. The result are disappointing as the average access time (got with JMeter) is around 2000ms. In JProfiler, I found that OGNL expressions compilation uses 40% of processing time. As the JMeter script is requesting always the same page on the webapp, I'm wondering why the expressions are always compiled. The cache seems not to be used: 31896 invocations on ExpressionCache.getCompiledExpression 31140 invocations on Ognl.compileExpression Is there someone to give me a clue to have better performances ? Thanks PS: sorry for double posting in dev list, I had some problems to subscribe to users. -- Denis Queffeulou - 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]
Re: T4.1.5 possible ognl cache problem
I've made another test with 4.1.5 which gives: 34,2% 90528 invocations on ExpressionCache.getCompiledExpression 25,4% 90300 invocations on Ognl.compileExpression The test is made of 5 pages which returns XML and has run for about 2 hours. I'm not sure to be able to test with 4.1.6 in the next days. -- Denis Jesse, could you comment on this? I haven't tried profiling my apps, but the repeated calls to Ognl.compileExpression are worrisome if Denis got his numbers right. Are there any ognl-related changes in 4.1.6? Denis, maybe you could repeat your profiling with the latest snapshot, so hopefully something could still be done for this before 4.1.6 is released. Kalle On Thu, Aug 28, 2008 at 5:08 AM, denis queffeulou < [EMAIL PROTECTED]> wrote: Hi all, I'm making some profiling on an web app made with Tapestry 4.1.5 which returns XML. The result are disappointing as the average access time (got with JMeter) is around 2000ms. In JProfiler, I found that OGNL expressions compilation uses 40% of processing time. As the JMeter script is requesting always the same page on the webapp, I'm wondering why the expressions are always compiled. The cache seems not to be used: 31896 invocations on ExpressionCache.getCompiledExpression 31140 invocations on Ognl.compileExpression Is there someone to give me a clue to have better performances ? Thanks PS: sorry for double posting in dev list, I had some problems to subscribe to users. -- Denis Queffeulou - 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]
T4.1.5 possible ognl cache problem
Hi all, I'm making some profiling on an web app made with Tapestry 4.1.5 which returns XML. The result are disappointing as the average access time (got with JMeter) is around 2000ms. In JProfiler, I found that OGNL expressions compilation uses 40% of processing time. As the JMeter script is requesting always the same page on the webapp, I'm wondering why the expressions are always compiled. The cache seems not to be used: 31896 invocations on ExpressionCache.getCompiledExpression 31140 invocations on Ognl.compileExpression Is there someone to give me a clue to have better performances ? Thanks PS: sorry for double posting in dev list, I had some problems to subscribe to users. -- Denis Queffeulou - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]