Slashdot Mirror

← Back to Stories (view on slashdot.org)

Build Distributed Object Mgmt Frameworks for J2EE

Posted by Hemos on from the for-those-few-who-need-it dept.

An anonymous reader writes "Many enterprise Java technology developers build their own object management infrastructures to improve application performance. However, traditional object pools encounter problems in applications that run across distributed JVMs on multiple physical machines. In this article, Zhengrong Tang presents an object management framework that uses the concept of scopes to handle distributed systems with ease."

8 comments

  1. I got out of bed... by Anonymous Coward · · Score: 0

    for this? Seriously, it's Monday morning, I need something interesting to look at. This, as practical as it may be, smacks *way* too much of work!

    1. Re:I got out of bed... by aflat362 · · Score: 1
      I need something interesting to look at

      Then you're on the wrong site. Boobies links are over at fark.

      --

      Conserve Oil, Recycle, Boycott Walmart

  2. Way too Heavy by Anonymous Coward · · Score: 0

    This article takes something simple, like displaying a web page and assumes that distributing a problem across JVM's is of the same complexity.

    The articles content is way too heavy for a simple site or medium site, and too naive for a high-end site.

    I would have more confidence if I saw some performance #'s to back up this work.

  3. A bit expensive... by tcopeland · · Score: 1

    ...it would take a pretty heavy object to outweigh the expense of the pooling outlined here - JMS message traffic, etc.

    But perhaps it would be handy for large documents or a connection to a database that was very slow or something like that...

    --
    The Army reading list
    1. Re:A bit expensive... by anomalous+cohort · · Score: 3, Insightful

      It's an old story. Users want both scalability and performance. So, they cache objects to improve performance but they also want clustering to scale up to larger volumns.

      This article is just speaking to that need for both. It's a hard road, by the way, as each node in the cluster must notify the other nodes when an object gets dirty. The hard part is doing that in a way that scales, is accurate, and without a prohibitive cost in message traffic.

    2. Re:A bit expensive... by tcopeland · · Score: 1

      > cache objects [...] but also want clustering

      So true. Talk about conflicting forces...

      > without a prohibitive cost
      > in message traffic

      Yup. JMS probably isn't a bad way to do this, especially as all the infrastructure classes have already been loaded into the VM as part of the J2EE server.

      Not the most popular Slashdot thread ever, huh? Nobody here but us chickens...

      --
      The Army reading list
  4. Deja vu heap layers by bdubSOv1iKIJ403M · · Score: 2, Informative

    This reminds me of Heap Layers and others; they both try to reinvent malloc()/new().