Visit Open-E website
Results 1 to 3 of 3

Thread: CPU versus Memory

  1. #1

    Default CPU versus Memory

    I have read (in this forum) about people using dual quad-core servers with tons of RAM.

    For Open-e, what provides a better return: CPU or RAM?
    Is there an optimum level, after which there is no appreciable gain?

  2. #2

    Default

    Hi

    I'm the guy who had loads of memory en dual quad 5335.
    After a few weeks of testing my findings about this are that open-e or the iscsi target that is used (iscsitarget)
    • does not use more than one CPU
    • and when using blockio memory is not used because filesystem cache is bypassed


    In all my tests i've never seen the cpu usage go above 100%, ( with 8 cpu's you can go to 800%) i've seen it.
    blockio is the current default and fileio is unusable because of initialization failing after 3 iscsi volumes.

    My advise is to go for with and
    • a fast dual single core machine
    • 2~4 GB of memory
    • a good raid controller ( i use a areca 1230ML 12 disk )
    • good ethernet controllers ( dual intel server e1000 )


    when reading large chunks of data from the open-e san i get the bonded dual e1000 full ( 200+ Mb/s)
    when writing large chunks of data to the open-e san I get 100Mb/s
    when reading & writing I get on 1 interface i get 100Mb in and out of it, using the full duplex

  3. #3

    Default

    Update of my previous post

    I've done some testing and i see that there is a difference in IOps when using 1 CPU instead of 8. There is a decrease in speed. my gues is 10 - 50 % drop. It depends on the size of the io size you are reading/writing.

    So multiple CPU's are beter then one. How many is useful is still being tested.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •