Re: [Ltsp-discuss] Raid 1 or Raid 5? New server advice.

[H. Sami Sozuer]

Ben Green wrote: On Mon, 21 Aug 2006 08:47:05 +0100, Gudmund Areskoug <[EMAIL PROTECTED]> wrote: Krsnendu Dasa wrote: Why are Opteron's recommended? On Anandtech, Tom's Hardware etc they all say that the new Xeon Dual Core beats the Opteron hands down. beats it under what circumstances? AFAIR, Opterons suitability for LTSP lies largely in how it handles multitasking/multithreading, so it might be slower on a certain kind of benchmark while remaining on top when it has to serve multiple users running many apps simultaneously. Or is the Xeon improved/better in that respect too? BR, Gudmund I would have thought that for LTSP, we'd want as many processors as possible, without any one processor being too slow. My reasoning is that more threads means more single threaded apps can be running without impinging on other apps, whilst you can still run big apps quickly. For example I recently had the choice of a quad pIII 750, or a dual Xeon 1.4. I went for the dual, as there would be more power for single apps on that set up, so heavy stuff like open office would run faster, rather than just running on one of the PIIIs. I am not sure if my conclusion are correct, and would love some concrete varification. I wonder how different numbers of concurrent users would affect this too. Would I have been better of with the quad for a larger number of users? I guess an important parameter in this regard would be the front bus speed. The Pentium/Xeon SMP architecture AFAIK share the same bus whereas the Opterons each have their own dedicated 1GHz FSB to their own dedicated memory, while each opteron can still access the other processors' memory, albeit at a cost of speed. Hence the name non uniform memory access, or NUMA. So for instance you'd have a total of roughly 4 GHz FSB for a quad opteron system whereas you're bound with 1066/1333 MHz FSB in a latest Xeon SMP box. So if the P4 750 and the Xeon system that you mention had the same FSB speed: You'd have been better off with the quad PIII if a large number of small threads that fit in the cache of each processor. I'd guess their performance would be about the same for a large number of threads requiring frequent memory access, and you'd be better off with the Xeon running larger applications. So it's a combination of processor speed, cache size, FSB and the number and memory access needs of the threads that together determine the overall performance. There is unfortunately no straightforward formula--pretty complicated stuff actually.

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_____________________________________________________________________
Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
      https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help,   try #ltsp channel on irc.freenode.net