Re: Dual-core CPU vs. very large cache



Shane Ambler wrote:
On 25/4/2006 22:37, "Bill Moran" <wmoran@xxxxxxxxxxxxxxxxxxxxxxx> wrote:

[First off, the list archives for this list don't seem to be
searchable. I get the following error:
Unable to read document excerpts
'/usr/local/mailman/archives/private/freebsd-performance/htdig/db.excerpts'
Did you run htdig?]

So ... on to the question.

We have some database servers that we're looking to replace with
beefier hardware, mainly because we're expecting our customer base
to grow a lot in the near future.

The current hw is Dell 2850 servers. These are dual proc (each proc
is hyperthreaded) with Dell PERC controllers driving 4 SCSI-320
disks in a RAID-10.

We're doing our best to simulate high-load in the lab, and the
database consistently bottlenecks on CPU usage. I'm assuming that
the combination of plenty of RAM and high-speed disks has led to
the CPU being the slowest part of the system.

We're considering two alternatives for the newer hardware:
1) Intel HT CPUs with 8M cache
2) Intel dual-core procs

Our current Dells have 2M cache, and I'm trying to determine whether
the 8M cache will make a significant difference or not. Can someone
recommend a testing procedure for determining whether adding cache is
worthwhile or not? I can simulate a test load at any time, but I
don't know how to tell whether the cache is the bottleneck of the
CPU or not.

Cache helps speed things up by keeping code/data in faster memory - this
helps speed things up when the same code/data is used repeatedly.

From the info you have given I am guessing that you have many users who are
loading the system up to capacity and that the database is fairly large
(a few Gig).
On that premise I would recommend the dual core CPU's (two or more dual core
xeon's - not P4's) - it sounds like they would be working with more data
than would be kept in cache so the extra cache wouldn't increase performance
a great deal and with many users loading the system the more cpu's of the
dual cores cpu's would allow more requests to be processed at the same time.

We've done extensive testing of dual-core systems here with different cpu bound processes, and have found out over and over that increasing the cache helps much more for cpu bound processes than dual core. If anything, get two single core processors with the 8M cache on them.

Also, make sure that your database is set up with indexes correctly and is pruned (if that needs to be done for your database server).


Eric




--
------------------------------------------------------------------------
Eric Anderson Sr. Systems Administrator Centaur Technology
Anything that works is better than anything that doesn't.
------------------------------------------------------------------------
_______________________________________________
freebsd-performance@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscribe@xxxxxxxxxxx"



Relevant Pages

  • Re: Dual-core CPU vs. very large cache
    ... The current hw is Dell 2850 servers. ... database consistently bottlenecks on CPU usage. ... Our current Dells have 2M cache, and I'm trying to determine whether ... On that premise I would recommend the dual core CPU's (two or more dual core ...
    (freebsd-performance)
  • Re: Purchasing the correct hardware: dual-core intel? Big cache?
    ... Dual core is faster than hyperthreaded ... CPU's usually about 20% if you use the larger CPU cache models. ... We're looking hard at getting either Intel dual-core procs, ...
    (freebsd-questions)
  • Opteron 165 for gaming. This a good idea ??
    ... I think you could overclock it to 2.5ghz easy. ... And the Dual core and ... the 2 meg cache should be good. ... Its a new cpu so not alot abour ir. ...
    (comp.sys.ibm.pc.games.action)
  • IPv6 oops on ifup in latest BK
    ... Using ACPI for SMP configuration information ... Initializing CPU#0 ... CPU: Trace cache: 12K uops, ...
    (Linux-Kernel)
  • SCSI CDROM issue in kernels >= 2.6.14-rc3
    ... CPU: Trace cache: 12K uops, ... MEM window: disabled. ... SCSI device sda: 17928698 512-byte hdwr sectors ...
    (Linux-Kernel)