Dear Cory,
my apologies for not being clear enough.
For the asset store, you're fine with slower disks tier 3 storage
products. But it's indeed more particular the postgres and lucene I/O load
that require fast disk access. You're totally right.
best regards,
Bram Luyten
On Thu, May 22, 2008 at 3:14 PM, Cory Snavely [EMAIL PROTECTED] wrote:
Our experience is markedly different, and I'm particularly struck by the
comment about 15Krpm disk.
We use 15Krpm SCSI disk in RAID 1 for Postgres, but we use SATA RAID 6
for the assetstore and do not see DSpace I/O-bound against it. (FYI
we're working on transitioning our assetstore to a Sun Honeycomb.)
I'm not sure your units are correct for the assetstore--7GB?--but if so,
sure, a little space on your 15Krpm disk is fine. If you mean 7TB, then
you're talking considerable expense over SATA that I am pretty confident
would *not* give any benefit. I/O against the assetstore you is low
volume sequential read and occasional high-volume sequential write, and
that is 100% consistent with tier 3 type storage products and what
SATA RAID does best. In fact I would think it would be perfectly
reasonable to put the assetstore on NAS.
FWIW we are currently deploying a replacement server for DSpace; it's a
dual quad-core Xeon server w/ 8GB RAM with 15Krpm internal SAS disk in
RAID 1 and 9TB SATA RAID 6. Our experience has shown that we need to
handle heavy multiprocessing Postgres load as well as large memory
allocation and that aside from Postgres storage I/O requirements are
relatively light.
Cory Snavely
University of Michigan Library IT Core Services
On Thu, 2008-05-22 at 14:54 +0200, Bram Luyten wrote:
Hello Jim,
we have installed and manage some very large DSpaces, but also a few
moderate ones. An example of a large installation:
http://lirias.kuleuven.be holds around 130.000 items, of which only
around 2500 of them contain full-text. This asset store is around 7GB.
It's mainly academic research output (papers, conference
presentations, ...). So currently, the average size of a bitstream is
around 2.8MB. But this will be very different if your repository is
oriented towards datasets, audio or video.
Concerning processing power and memory: the system currently has
around 1000 unique visitors daily. There are 4000 e-persons. During
office hours, we experience an average of 4 concurrent logged-in
users, performing submissions, etc (= intensive on database and
indexes). The tomcat has been given 2GB of memory, while the whole
system has 3.5GB of RAM. This doesn't really cover the total load, as
we use swapping, but it's enough to keep the system running. A
recommendation: don't cut down on disk speed, you will need 15.000 rpm
disks.
The server has 1 physical dual-core processor, with hyper threading
(so actually 4 virtual processors). In peak times, this is becoming a
bottle neck.
If you could illustrate the purposes of your installation, or the
estimated number of users, I could provide you with a specific case,
that possibly more closly matches what you're looking at.
with kindest regards,
Bram Luyten
--
@mire NV
Romeinse Straat 18
3001 Heverlee
Belgium
+32 2 888 29 56
http://www.atmire.com - Institutional Repository Solutions
http://www.togather.eu - Before getting together, get [EMAIL PROTECTED]
On Wed, May 21, 2008 at 4:33 PM, Jim Price
[EMAIL PROTECTED] wrote:
Hello,
We are currently running a test instance on a Sun Enterprise
450 running Solaris 10. We are currently looking into new
hardware for a production environment.
We would like to find out if anyone is willing to share their
experiences with hardware?
What platform would you recommend for a school starting with
dspace?
What are you using for hardware?
What do you store?
How big is your data storage?
How much growth did you see in your first year of use?
What kind of loads are you experiencing?
Thanks,
Jim
-
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/
___
DSpace-tech mailing list
DSpace-tech@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dspace-tech
-
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/
___ DSpace-tech mailing list
DSpace-tech@lists.sourceforge.net
