Version 18

    Clustering Projects


    1. JBoss AS

      1. Clustering component in JIRA

      2. Contacts: Dimitris Andreadis (lead, Greece), Scott Stark (guru/profile service/deployers, Washington), Adrian Brock (guru/microcontainer/deployers, Neuchatel), Ales Justin (microcontainer, Slovenia)

    2. EJB3

      1. Clustering component in JIRA

      2. Has morphed into being key player in SFSB caching in general

      3. Contacts: Carlo de Wolf (lead, Holland); Andrew Rubinger (NYC/Boston)

    3. Hibernate Core

      1. cache-jbosscache subproject

        1. JBC 2 as a Hibernate 2nd Level Cache

      2. Contact: Steve Ebersole (lead, Austin)

    4. JBoss Messaging

      1. Coordinating usage of JGroups in JBoss AS

      2. No real code involvement

      3. Contacts: Tim Fox (lead, UK); Clebert Suconic (Austin)

    5. JBoss Cache

      1. less code involvement than in past; they have more resources, we've got too much to do

        1. hopefully that will change in a year or so as we clear our roadmap

      2. lots of design discussion involvement, user forum participation

      3. Contacts: Manik Surtani (lead, UK); Mircea Markus (Romania)

    6. PojoCache

      1. little code involvement, some design discussion involvement

      2. Contacts: Jason Greene (lead, Madison)

    7. JGroups

      1. Involvement is similar to JBC

      2. Multiplexer vs. Shared Transport:

      3. Contacts: Bela Ban (lead, overall clustering lead, Switzerland), Vladimir Blagojevic (Toronto)

    8. jboss-ha libraries

      1. jboss-ha-client and jboss-ha-server-api

        1. libraries that can be consumed by both JBAS and EJB3

      2. Managed JBCLUSTER project in JIRA

        1. project is quite confused; also used for high level roadmap


    AS Clustering Major Issues


    1. Web Tier Clustering

      1. 5.0.0.CR?

        1. Scoping of FIELD granularity pojos


          2. Scope to the session rather than the webapp

          3. Add JBC Region per session, remove in invalidate

          4. clean out junk that resulted from per webapp scoping

            1. Observer mixin

            2. Deploy std PojoCache jboss-aop.xml

            3. Pojo eviction configuration

            4. need new jboss-metadata release

        2. Determine how to clean up persistent store around undeploy/deploy

          1. Check perf impact using region-based marshalling

      2. mod_cluster


        2. One of Brian's 2 priorities for summer

      3. Possible cooperation with Mobicents

        1. Remind me to forward email threads

      4. Sharable web session


        2. Longstanding feature request w/ a lot of votes

      5. Clustered ServletContext (replicated attribute map)


      6. Clean shutdown problem?? (Update: Looks like Remy fixed this --

    2. EJB3 SFSB Clustering

      1. 5.0.0.CR2

        1. Clean shutdown valve


      2. (Paul) New caching implementation


        2. Would love to see this in 5.0.0; planned for 5.1.0.

      3. Long term: FIELD Granularity

    3. EJB3 Entity Clustering

      1. 5.0.0.CRX

        1. (Paul next week???) Integrate my new Hibernate/JBC2 caching impl

          1. Determine what to do with old implementation

          2. Still ship/support in EJB3? (yuck!!)

          3. Extract and preserve for Hibernate 3.2.6 users

      2. Benchmarking (lower priority)

    4. (Paul) Coordinated replication -- Seam case

      1. HttpSession + SFSB + JPA




      2. Coordination of cache writes tx and request end

      3. Optimized serialization

    5. Clustered Deployments (2nd of Brian's priorities for summer)

      1. 5.0.0.CR2

        1. HASingleton deployer replacement



        2. Clustered DeploymentRepository (replaces AS 4.x FarmService)


        3. Add HAPartition and CacheManager dependencies to clustered EJBs

      2. Clustered DeploymentManager

        1. server side

        2. client side in embedded console + JON

      3. Rationalize clustering metadata


    6. Messaging

      1. 5.0.0.CR2

        1. Use same shared transport for JBM and other AS UDP channels


        2. Benchmark effect of enable_bundling on HTTP session replication


    7. HAPartition

      1. 5.0.0.CR2

        1. Confirm that it properly handles concurrent invocations from JGroups (preferably with tests


    8. Quality

      1. 5.0.0.GA

        1. (Dominik) Testing of multiple-service apps (aka Seam use case)

        2. (Dominik) Network failure testing

        3. Testing of concurrent calls into HAPartition

        4. Soak testing

          1. HttpSession

          2. EJB3 SFSB

      2. War on JBC regressions

        1. Add tests in JBC testsuite that specifically exercise AS/EJB3/Hibernate call patterns using configurations that mirror the ones we use

      3. Refactor to allow shift away from overly heavy reliance on integration testing

        1. Unit and subcomponent tests

    9. Docs

      1. 5.0.0.GA

        1. Need to figure out how to involve Samson Kittoli as much as possible

        2. Usage of JGroups ChannelFactory

        3. Usage of JGroups shared transport

        4. Usage of JBC CacheManager

          1. effect on jboss-web.xml and EJB3 SFSB annotations, Clustered SSO

        5. Buddy replication!!

        6. Clustered HttpSession passivation

    10. Management

      1. Add clustering management to the embedded console / JON

      2. Critical weakness!!!

    11. Network partitions

      1. Primary partition

        1. Notifications, pluggable policies to allow services/servers to shutdown when their view has less than X members

      2. State merging (lower priority)

    12. Benchmarking

      1. Need a program of consistently runnable benchmarks, currently too manual

        1. Expand on what Dominik's done

        2. Need more scenarios: granularities, triggers, expiration/invalidation

      2. Need EJB3 SFSB as well

    13. Misc

      1. (Paul) Clustered EJB timers