Re: [ldm-users] LDM: Hey, hey, you, you, get onto my cloud

• To: "ldm-users@xxxxxxxxxxxxxxxx" <ldm-users@xxxxxxxxxxxxxxxx>
• Subject: Re: [ldm-users] LDM: Hey, hey, you, you, get onto my cloud
• From: "Wojtowicz, David P Jr" <davidw@xxxxxxxxxxxx>
• Date: Fri, 17 Jun 2011 23:14:05 +0000

Virtualization guys rarely understand full-load machines (like a busy LDM
server).  My guess is the virtual machine won't be able to keep up with
demand at first on the initial settings.  But rather than be embarrassed
after talking you into it, they'll have to provision a much larger
dedicated chunk of their resources than they anticipated it costing them.
You come out the winner there with expensive hardware being essentially
dedicated to serving you but provided by someone else.

We're virtualizing more and more stuff here, and it can provide some
dramatic efficiencies, but only for certain kinds of loads.  Virtualizing
a 100% busy machine costs more than operating a dedicated server.

I'd certainly run the queue in RAM instead of a disk file.  We've been
using RAM-only queues for years on various LDM servers and have gotten
away with using less powerful IO than would be needed otherwise.

  
----

David Wojtowicz, Coordinator of Research and IT
School of Earth, Society and Environment (SESE)
University of Illinois at Urbana-Champaign
davidw@xxxxxxxxxxxx   +1 217 333 8390

• Follow-Ups:
  • Re: [ldm-users] LDM: Hey, hey, you, you, get onto my cloud
    • From: Gilbert Sebenste