Re: Caching in a cluster - Weblogic

This is a discussion on Re: Caching in a cluster - Weblogic ; Marco, > i have written a servlet that acts as a cache for database data, since access > to that data takes time and EJB entity was not considered a proper solution since > our apps need just to read ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: Re: Caching in a cluster

  1. Re: Caching in a cluster

    Marco,

    > i have written a servlet that acts as a cache for database data, since

    access
    > to that data takes time and EJB entity was not considered a proper

    solution since
    > our apps need just to read data from Database.
    > so we decided to opt for a servlet that will have internally a cache of

    the DB
    > data, and all the apps that need to retrieve data from DB will contact the

    servlet
    > via HTTP.
    > Now, issue is that in its init() method the servlet loads the data from

    the database,
    > and if i deploy the servlet in the clustered environment mentioned above,

    how
    > many times the servlet will be initialized? only once since it is in a

    cluster
    > or one time per server (and thus 6)?


    For each server that loads the app, the servlet will be initialized on that
    server. (With some recent versions of WebLogic with 2-stage deploy, you may
    even see it initialized twice on each server.)

    To do read-only caching, you want each one to load the data (so using the
    init() is OK).

    For clustered caching of data that can change, use Coherence:
    http://www.tangosol.com/coherence.jsp

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    http://www.tangosol.com/coherence.jsp
    Tangosol Coherence: Clustered Replicated Cache for Weblogic


    "Marco" wrote in message
    news:3f13a620$1@newsgroups.bea.com...
    >




  2. Re: Caching in a cluster


    "Cameron Purdy" wrote in message
    news:3f1b5515@newsgroups.bea.com...
    > For clustered caching of data that can change, use Coherence:
    > http://www.tangosol.com/coherence.jsp



    Or of course, any of a number of free, open source implementations based on
    the excellent Java Groups.
    See:
    JavaGroups: www.javagroups.com
    OSCache - http://www.opensymphony.com/oscache/clustering.html
    SwarmCache - http://swarmcache.sourceforge.net/

    Regards,
    Jon



  3. Re: Caching in a cluster

    "Jon" wrote in message news:<3f1c284c@newsgroups.bea.com>...
    >
    > Or of course, any of a number of free, open source implementations based on
    > the excellent Java Groups.


    Been there, done that.. Java Groups is slow and worse.. it's buggy and
    eats loads of memory when there is any traffic. JBOSS uses Java Groups
    for clustering and it's a nightmare to scale past a couple servers.
    Their docs (pay for subscription) suggest limit the clusters to 2 or 3
    servers with a load balancer to spread the load across lots of little
    clusters..

    -sudjutin

  4. Re: Caching in a cluster

    Hi Sudjutin,

    I experienced nothing of the sort. In fact, I find it works quite well...I
    have also found the developers pretty responsive to problems. If you have a
    problem, bring it up...

    Jon

    wrote in message
    news:5dfc30a1.0307211921.7d4fbfdf@posting.google.c om...
    > "Jon" wrote in message

    news:<3f1c284c@newsgroups.bea.com>...
    > >
    > > Or of course, any of a number of free, open source implementations based

    on
    > > the excellent Java Groups.

    >
    > Been there, done that.. Java Groups is slow and worse.. it's buggy and
    > eats loads of memory when there is any traffic. JBOSS uses Java Groups
    > for clustering and it's a nightmare to scale past a couple servers.
    > Their docs (pay for subscription) suggest limit the clusters to 2 or 3
    > servers with a load balancer to spread the load across lots of little
    > clusters..
    >
    > -sudjutin




  5. Re: Caching in a cluster

    sudjutin@amuro.net wrote in message

    > Been there, done that.. Java Groups is slow and worse.. it's buggy and
    > eats loads of memory when there is any traffic.


    [I was pointed to this thread by someone who subscribes to this
    newsgroup]

    I think JavaGroups is great ! But then again, I'm a little bit biased
    because I wrote it :-)

    Can you tell me what exactly is slow with JavaGroups ? This is a
    pretty vague statement, and since JavaGroups can be configured to
    death, there are many situations where misconfiguration results in
    slow (and possibly incorrect) behavior.

    We ship a number of standard and well-tested stack configurations.

    I have just recently conducted tests with 4 nodes and 2 senders, and
    sending 1.4 billion (10K) messages, and got an average sustained
    throughput of 3MB/sec. Increasing the msg size to a number just below
    the UDP frag size would probably increase, possibly double, the
    throughput. During the tests, the max resident size of each node was
    35MB.

    I'll shortly post an article on this; comparing JavaGroups to Spread
    and JBora.


    > JBOSS uses Java Groups
    > for clustering and it's a nightmare to scale past a couple servers.


    We have customers who have happily scaled past 10 nodes.

    Bela

  6. Re: Caching in a cluster

    > What kind of hardware was that run on? (Servers, network speed, switch type,
    > etc.) What was the CPU utilization of the boxes? Any idea on the network
    > utilization?



    4 nodes (2 senders). Each machine was a SunBlade 1000, 512MB mem,
    750MHz. LAN was a 100MB switched eithernet. Load was during 4th of
    July weekend, so light.

    I get the 3MB/sec sustained throughput also during business hours,
    when the CPU load and the nw util are higher (only for 2 million msgs
    though, not 1.4 billion).

    Bela

+ Reply to Thread