Discussions

Blogs: Henrik Stahl: How to get (almost) 3 GB heap on Windows

  1. Henrik Stahl has posted a blog entry on how to get 3 GB heap on Windows with the BEA JRockit VM. While that is impressive in and of itself (given Windows' 2GB process limit) the grand questions then become: is this really a good thing? How many processes need this much RAM, and why don't any of my machines have more than 2GB onboard? :)

    Threaded Messages (6)

  2. You can now get 128GB of RAM on a commodity server (3U, 4xSocket, 8xCore, 8xDIMM per socket). Most 2-socket (dual CPU) servers are coming with 4-8GB standard now. Even 1-socket (single CPU) servers are coming with 4GB.

    While we don't see a ton of Windows on the server, we're definitely seeing a lot of commodity servers being used to manage large amounts of data in memory (cache servers, data grid nodes, etc.) So this is good news.

    Peace,

    Cameron Purdy
    Tangosol Coherence: Clustered Shared Memory for Java
  3. More RAM better[ Go to top ]

    I'll second cameron's comment. At work, we're already hitting the 2Gb limit. Many of the servers I'm using these days come with 4Gb of RAM, so using 3Gb heap sounds good to me.

    For things like rule engines and heavy applications, 2Gb is rather tiny. This is in a real-time context :)

    peter
  4. Are you seeing people move to the 64 bit JVMs too ?
  5. Migration to 64-bit JVMs[ Go to top ]

    We see growing interest in 64-bit JVMs, but most seem to start by upgrading their OSes to 64-bit and keep running 32-bit JVMs. The two major hurdles for moving to 64 bits are 1) performance is typically worse and 2) compatibility issues with 32-bit native libraries.

    We are working on various ways of alleviating these issues. Stay tuned...

    Henrik
    BEA JRockit
  6. What about UTF8 format compatibility..[ Go to top ]

    64 bit JVM's sounds nice, but how about UTF8 format compatibility issues??
    How's that going to be resolved???
  7. What about UTF8 format compatibility..[ Go to top ]

    64 bit JVM's sounds nice, but how about UTF8 format compatibility issues??How's that going to be resolved???

    UTF8 is a binary character encoding, and it can be used in any 16-bit, 32-bit or 64-bit system.

    The only problem that I'm aware of with Java's implementation is that the java.io.DataOutput interface defines a binary format that limits UTF8 values to 65,535 characters in length. That has nothing to do with 32-bit or 64-bit JVMs, though.

    Peace,

    Cameron Purdy
    Tangosol Coherence: Clustered Shared Memory for Java