would be the same, @Lock don't forbid you to use @Lock(WRITE) later but tha's up to you if you don't need it
Romain Manni-Bucau Twitter: @rmannibucau Blog: http://rmannibucau.wordpress.com/ LinkedIn: http://fr.linkedin.com/in/rmannibucau Github: https://github.com/rmannibucau 2014-09-04 10:55 GMT+02:00 Lars-Fredrik Smedberg <[email protected]>: > @Romain > > Or should I even do @Singleton @ConcurrencyManagement(BEAN) if I need no > state? I know you said that the performance implications would be about the > same? > > > On Thu, Sep 4, 2014 at 10:39 AM, Romain Manni-Bucau <[email protected]> > wrote: > >> hmm for webservices which are most of the time staeless I'd use >> @Singleton @Lock(READ), do you need any state? >> >> >> Romain Manni-Bucau >> Twitter: @rmannibucau >> Blog: http://rmannibucau.wordpress.com/ >> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> Github: https://github.com/rmannibucau >> >> >> 2014-09-04 10:30 GMT+02:00 Lars-Fredrik Smedberg <[email protected]>: >> > @Alex and @Romain >> > >> > Thanks for your answers. I forgot to mention that we need (for most of >> the >> > cases) transactions. >> > >> > For some of the production environments we use WebSphere 8.5.5 which >> still >> > is JavaEE 6 so we need to use EJB to handle transactions. >> > >> > @Romain >> > >> > I agree with your thoughts, the requirements from case to case should >> > decide the inplementation. I just thought listing pros and cons was a >> good >> > way to sort of wrap our heads around what might be things to consider. We >> > expose services using JAX-WS and/or JAX-RS and we need transactions. >> > Considering that I understand that you rather would use @Stateful >> > @RequestScoped rather than @Stateless? >> > >> > @Alex >> > >> > I see your comment on the fact that multiple requests each will use its >> own >> > instance of a stateless EJB. i assume however that @PostConstruct will >> only >> > be called once when the container instantiate the EJB and put it in the >> > pool and not once per request. So if I need request/client specific state >> > setup in @PostConstruct I would need to use a @Stateless @RequestScoped >> EJB >> > instead. Was your comment about that? >> > >> > Thanks >> > On 4 Sep 2014 09:22, "Alex Soto" <[email protected]> wrote: >> > >> >> BTW in Java EE 7 with @Transactional then you can create a POJO with >> >> RequestScoped and Transactional so no EJB is required. >> >> >> >> Moreover keep in mind that an stateless bean is used during the whole >> >> request, two concurrent requests won't reuse the same stateless bean. >> >> Alex. >> >> >> >> >> >> 2014-09-04 9:17 GMT+02:00 Romain Manni-Bucau <[email protected]>: >> >> >> >> > Hi >> >> > >> >> > interesting analyzis and way of doing it :). Personally I think the >> >> > other way around actually: >> >> > >> >> > 1) what do I need? >> >> > [potential answer] storing my state during JSF request -> I add >> >> > @RequestScoped >> >> > >> >> > 2) oops, I need transactions >> >> > -> add @Stateful (or use a service to delegate depending the case and >> >> > code architecture) >> >> > >> >> > etc... >> >> > >> >> > >> >> > What you forgot in pro/cons is stateless are pooled so they can be a >> >> > bottleneck if not well configured >> >> > >> >> > >> >> > Generally I don't use stateless anymore, only @Singleton for EJBs and >> >> > @Stateful if really a CDI bean doesn't match my case - for JSF you >> >> > often needs it to store a state within a scope (request, session, view >> >> > typically) and flush it at the end of the action in a transaction. >> >> > >> >> > So to come back to your question: start with the minimum you need and >> >> > don't try to get a "always use XXX", it would just be broken as all >> >> > general rules ;). Passing from an EJB to a CDI bean is almost nothing >> >> > to do and the only relevant info is a benchmark on *your* app (I saw >> >> > cases where postconstruct can be considered as free and cases where >> >> > injections were costly, so it depends too much on the code you >> >> > evaluate to be general. >> >> > >> >> > Hope it helps even if adding some blur ;) >> >> > >> >> > >> >> > >> >> > >> >> > >> >> > Romain Manni-Bucau >> >> > Twitter: @rmannibucau >> >> > Blog: http://rmannibucau.wordpress.com/ >> >> > LinkedIn: http://fr.linkedin.com/in/rmannibucau >> >> > Github: https://github.com/rmannibucau >> >> > >> >> > >> >> > 2014-09-04 8:48 GMT+02:00 Lars-Fredrik Smedberg <[email protected]>: >> >> > > Hi >> >> > > >> >> > > Trying to sort out the pros and cons of using a @Stateless EJB vs a >> >> > > @Stateful @RequestScoped EJB, can someone help me with the following >> >> > claims >> >> > > and see if I missed out on something?? >> >> > > >> >> > > Possible pros of @Stateless EJB >> >> > > >> >> > > - Any expensive @PostConstruct methods will be run once when the >> EJB is >> >> > > instantiated and put in the pool and not per request >> >> > > - Control over concurrency/resources on the server by being able to >> >> > > configure the EJB pools >> >> > > - Reuse of EJBs will not create much garbage to be handled by the GC >> >> > > >> >> > > Possible cons of @Stateless EJB >> >> > > >> >> > > - Having to tune and follow up on the pool usage. This might be >> >> > troublesome >> >> > > depending on the organization and responsibilities >> >> > > - Depending on the tuning requests might need to wait for an >> instance >> >> to >> >> > be >> >> > > available in the pool >> >> > > - @PostConstruct can not initialize any request/user dependent >> state of >> >> > the >> >> > > EJB since it will be shared by others >> >> > > >> >> > > Possible pros of @Stateful @RequestScoped EJB >> >> > > >> >> > > - No configuration of EJB pools needed >> >> > > - No calls will have to wait for an available instance, as many >> >> instances >> >> > > as needed will be created >> >> > > - @PostConstruct can be used to initialize request/user dependent >> state >> >> > of >> >> > > the EJB, the EJB instance will not be reused >> >> > > >> >> > > Possible cons of @Stateful @RequestScoped EJB >> >> > > >> >> > > - An expensive @PostConstruct will affect performance since it will >> run >> >> > > once per request (client request) >> >> > > - Generates more garbage to be handled by the GC >> >> > > - No built in way to control the number of concurrent calls >> >> > > >> >> > > >> >> > > - Are there any other obvious pros and cons with the two ways above? >> >> > > - In the end will it all depend on the use-case and how many pros >> the >> >> > > use-case/scenario will make use of? >> >> > > - Letting CDI lifecycle create and destroy stateful EJBs per request >> >> > > compared to handling a pool of stateless EJBs, is there a big >> >> performance >> >> > > difference? >> >> > > >> >> > > >> >> > > Hoping for help to shed some light on this >> >> > > >> >> > > Regards >> >> > > Lars-Fredrik >> >> > > >> >> > > >> >> > > -- >> >> > > Med vänlig hälsning / Best regards >> >> > > >> >> > > Lars-Fredrik Smedberg >> >> > > >> >> > > STATEMENT OF CONFIDENTIALITY: >> >> > > The information contained in this electronic message and any >> >> > > attachments to this message are intended for the exclusive use of >> the >> >> > > address(es) and may contain confidential or privileged information. >> If >> >> > > you are not the intended recipient, please notify Lars-Fredrik >> Smedberg >> >> > > immediately at [email protected], and destroy all copies of this >> >> > > message and any attachments. >> >> > >> >> >> >> >> >> >> >> -- >> >> +----------------------------------------------------------+ >> >> Alex Soto Bueno - Computer Engineer >> >> www.lordofthejars.com >> >> +----------------------------------------------------------+ >> >> >> > > > > -- > Med vänlig hälsning / Best regards > > Lars-Fredrik Smedberg > > STATEMENT OF CONFIDENTIALITY: > The information contained in this electronic message and any > attachments to this message are intended for the exclusive use of the > address(es) and may contain confidential or privileged information. If > you are not the intended recipient, please notify Lars-Fredrik Smedberg > immediately at [email protected], and destroy all copies of this > message and any attachments.
