Re: Performance and scalability
Until someone in the community (either David, or me, or indeed you) has done some formal performance/scalability testing, none of us can really answer this. A couple of years ago I was on a (non Isis) project where we used grinder [1],[2] for the performance testing. That worked well enough in that case, and I'd definitely think about using it for testing Isis. In the meantime you could do some googling on performance with the Wicket framework, and on performance of DataNucleus. Any issues/limitations will be an issue for Isis too. With Restful Objects, I think that will scale as far as you need because it is intrinsically stateless. With the Wicket viewer, it has a heavy dependency (through Wicket) on sessions, which will complicate matters if you find you need a web farm, cf [3]. HTH Dan [1] http://grinder.sourceforge.net/ [2] http://stackoverflow.com/questions/1209930/jmeter-versus-the-grinder [3] http://stackoverflow.com/questions/2255729/jsf-sessions-in-a-web-farm On 17 August 2013 22:04, james agada okwuiag...@gmail.com wrote: You misunderstand me. I am impressed with the framework and want to use it. We are currently doing a PoC. But I want to be able to tune it to support large number of concurrent users and I will appreciate advise on how to scale to thousands of users. To give an example, by using a mobile viewer, I can quickly build a class of apps that can gain thousands of users. How do I scale Isis to backend this? Is that even possible? Does it just scale like regular java web app? What kind of servers do I need? That is my agenda. Sent from my iPhone On Aug 17, 2013, at 9:58 PM, David Tildesley davo...@yahoo.co.nz wrote: I'm not sure what your agenda is Mr Agada, however I can assure you that, from our experience ISIS is not just a prototyping tool and we have put our application that uses ISIS through informal performance testing and we are confident. It is just that our application release one is batch driven use cases (which by the way have excellent performance driving updates through the ISIS DOM based domain logic) but only one (wicket viewer) human user (an admin), while our release two is planned to have a concurrency of 200 (wicket viewer) users with a high amount of query traffic compared with update traffic where we will be doing formal performance testing (which by the way is an expensive exercise). You should know that even very large vendors don't publish performance figures for their frameworks and certainly don't publish performance comparison benchmarks with competing frameworks, commercial or otherwise. Performance figures can be skewed by favoring certain use cases over others. I suggest you adopt a less accusatory tone and do your own performance benchmarking because at the end of day that is the only way you are going to answer your question to your own satisfaction. David. From: james agada okwuiag...@gmail.com To: users@isis.apache.org users@isis.apache.org Sent: Saturday, 17 August 2013 9:06 PM Subject: Re: Performance and scalability It should be on your priority except Isis will just be a prototyping tool. Sent from my iPhone On Aug 17, 2013, at 1:31 AM, David Tildesley davo...@yahoo.co.nz wrote: Performance testing our ISIS (RO, Wicket viewer) based application is on our backlog to do soon, however we haven't started yet - mainly because it is not a significant risk for our first release. Regards, David. From: Dan Haywood d...@haywood-associates.co.uk To: users users@isis.apache.org Sent: Saturday, 17 August 2013 3:08 AM Subject: Re: Performance and scalability Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Re: Performance and scalability
Thanks. I'll keep community posted as we progress. Sent from my iPhone On Aug 18, 2013, at 8:07 AM, Dan Haywood d...@haywood-associates.co.uk wrote: Until someone in the community (either David, or me, or indeed you) has done some formal performance/scalability testing, none of us can really answer this. A couple of years ago I was on a (non Isis) project where we used grinder [1],[2] for the performance testing. That worked well enough in that case, and I'd definitely think about using it for testing Isis. In the meantime you could do some googling on performance with the Wicket framework, and on performance of DataNucleus. Any issues/limitations will be an issue for Isis too. With Restful Objects, I think that will scale as far as you need because it is intrinsically stateless. With the Wicket viewer, it has a heavy dependency (through Wicket) on sessions, which will complicate matters if you find you need a web farm, cf [3]. HTH Dan [1] http://grinder.sourceforge.net/ [2] http://stackoverflow.com/questions/1209930/jmeter-versus-the-grinder [3] http://stackoverflow.com/questions/2255729/jsf-sessions-in-a-web-farm On 17 August 2013 22:04, james agada okwuiag...@gmail.com wrote: You misunderstand me. I am impressed with the framework and want to use it. We are currently doing a PoC. But I want to be able to tune it to support large number of concurrent users and I will appreciate advise on how to scale to thousands of users. To give an example, by using a mobile viewer, I can quickly build a class of apps that can gain thousands of users. How do I scale Isis to backend this? Is that even possible? Does it just scale like regular java web app? What kind of servers do I need? That is my agenda. Sent from my iPhone On Aug 17, 2013, at 9:58 PM, David Tildesley davo...@yahoo.co.nz wrote: I'm not sure what your agenda is Mr Agada, however I can assure you that, from our experience ISIS is not just a prototyping tool and we have put our application that uses ISIS through informal performance testing and we are confident. It is just that our application release one is batch driven use cases (which by the way have excellent performance driving updates through the ISIS DOM based domain logic) but only one (wicket viewer) human user (an admin), while our release two is planned to have a concurrency of 200 (wicket viewer) users with a high amount of query traffic compared with update traffic where we will be doing formal performance testing (which by the way is an expensive exercise). You should know that even very large vendors don't publish performance figures for their frameworks and certainly don't publish performance comparison benchmarks with competing frameworks, commercial or otherwise. Performance figures can be skewed by favoring certain use cases over others. I suggest you adopt a less accusatory tone and do your own performance benchmarking because at the end of day that is the only way you are going to answer your question to your own satisfaction. David. From: james agada okwuiag...@gmail.com To: users@isis.apache.org users@isis.apache.org Sent: Saturday, 17 August 2013 9:06 PM Subject: Re: Performance and scalability It should be on your priority except Isis will just be a prototyping tool. Sent from my iPhone On Aug 17, 2013, at 1:31 AM, David Tildesley davo...@yahoo.co.nz wrote: Performance testing our ISIS (RO, Wicket viewer) based application is on our backlog to do soon, however we haven't started yet - mainly because it is not a significant risk for our first release. Regards, David. From: Dan Haywood d...@haywood-associates.co.uk To: users users@isis.apache.org Sent: Saturday, 17 August 2013 3:08 AM Subject: Re: Performance and scalability Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Re: Performance and scalability
It should be on your priority except Isis will just be a prototyping tool. Sent from my iPhone On Aug 17, 2013, at 1:31 AM, David Tildesley davo...@yahoo.co.nz wrote: Performance testing our ISIS (RO, Wicket viewer) based application is on our backlog to do soon, however we haven't started yet - mainly because it is not a significant risk for our first release. Regards, David. From: Dan Haywood d...@haywood-associates.co.uk To: users users@isis.apache.org Sent: Saturday, 17 August 2013 3:08 AM Subject: Re: Performance and scalability Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Re: Performance and scalability
I'm not sure what your agenda is Mr Agada, however I can assure you that, from our experience ISIS is not just a prototyping tool and we have put our application that uses ISIS through informal performance testing and we are confident. It is just that our application release one is batch driven use cases (which by the way have excellent performance driving updates through the ISIS DOM based domain logic) but only one (wicket viewer) human user (an admin), while our release two is planned to have a concurrency of 200 (wicket viewer) users with a high amount of query traffic compared with update traffic where we will be doing formal performance testing (which by the way is an expensive exercise). You should know that even very large vendors don't publish performance figures for their frameworks and certainly don't publish performance comparison benchmarks with competing frameworks, commercial or otherwise. Performance figures can be skewed by favoring certain use cases over others. I suggest you adopt a less accusatory tone and do your own performance benchmarking because at the end of day that is the only way you are going to answer your question to your own satisfaction. David. From: james agada okwuiag...@gmail.com To: users@isis.apache.org users@isis.apache.org Sent: Saturday, 17 August 2013 9:06 PM Subject: Re: Performance and scalability It should be on your priority except Isis will just be a prototyping tool. Sent from my iPhone On Aug 17, 2013, at 1:31 AM, David Tildesley davo...@yahoo.co.nz wrote: Performance testing our ISIS (RO, Wicket viewer) based application is on our backlog to do soon, however we haven't started yet - mainly because it is not a significant risk for our first release. Regards, David. From: Dan Haywood d...@haywood-associates.co.uk To: users users@isis.apache.org Sent: Saturday, 17 August 2013 3:08 AM Subject: Re: Performance and scalability Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Performance and scalability
Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Re: Performance and scalability
Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?
Re: Performance and scalability
Performance testing our ISIS (RO, Wicket viewer) based application is on our backlog to do soon, however we haven't started yet - mainly because it is not a significant risk for our first release. Regards, David. From: Dan Haywood d...@haywood-associates.co.uk To: users users@isis.apache.org Sent: Saturday, 17 August 2013 3:08 AM Subject: Re: Performance and scalability Not yet. It is something I intend to do in developing Estatio, though it's possible that others here (David?), might do look into this first. I would expect that the RO viewer would scale further than the Wicket viewer, because the former is stateless, the latter (due to Wicket's architecture) is not. Dan On 16 August 2013 16:06, james agada okwuiag...@gmail.com wrote: Does anyone have any experience and/or benchmarks and/or guides for Isis performance and scalability with wicketviewer or with the restfulobject viewer?