articles
FORUMS!
links
contact

PC Database
([an error occurred while processing this directive] entries)

Folding Team

SETI@HOME Team
RC5 Team
Genome Team

OCAU chat

SAN Benchmarks: What do you really get for $300k?
To show you the performance of this storage I commandeered a Compaq Proliant DL360 (Or Pizzabox as I call them) with Dual P3 1GHz CPU’s and 2.25GB of RAM. For anyone in Corporate IT, I can’t say enough good things about these little systems. Good price, small space, great performance. Great for a SAN environment where storage is external to the servers.

click to enlarge - pizzaboxes

I opted to use the DL360 as a Dual P3 is closer to the desktop systems that most Overclockers have. Alternately I could have used one of our Proliant 6400R Quad P3 Xeon Systems as shown here. Just imagine what this rack could do for OCAU’s folding team.

click to enlarge - quad xeons

Anyway, so I threw together a small RAID 0 disk array to run some benchmarks on:

click to enlarge - the OCAU array

Ok, so a 200GB array consisting of 12 x 18GB 10,000RPM drives isn’t exactly small, but we’re obscenely rich remember?

I initially tried to benchmark the disk with SiSoft Sandra, but the benchmarks rated the disk as slower than a couple of IDE drives. The problem is that Sandra obviously wasn’t designed with this sort of thing in mind and just wasn’t pushing the disk hard enough. HDTach's free version only works with Win98 and we're using Win2k of course. So I moved to Intel’s Iometer and ran tests again. This produced the expected results and allows us to cross reference with figures borrowed from the www.storagereview.com database. As has been said on the news page a few times on OCAU, StorageReview are a great site for HDD info.

Before I give you a couple of numbers, keep in mind that the tests were run during normal business hours. In other words the Servers, SAN, RAID controllers and other drives were all in full use, so this isn’t a sterile test environment designed to give maximum figures. Also, keep in mind these disks are 600 meters away from the server at the end of a fibre optic cable. Let’s see you do that with ATA100. ;) Also, this is not the "burst" or straight data-pumping figure you might be used to seeing elsewhere. This benchmark simulates real fileserver activity.

Intel's IOMeter Benchmark - IO/sec

Intel's IOMeter Benchmark - MB/sec

Intel's IOMeter Benchmark - Response Time (lower is better)

For those that prefer some real world examples, it takes 1:31 to copy 1.76GB from one directory on the drive to another. Not bad for a simultaneous read/write operation. In the 64K streaming read Iometer test, we managed a peak read of 48.67MB/s using a 32bit/33Mhz PCI card in the server. We also have 64bit/66Mhz cards, which weren’t available for testing at the time. This should give a bit more bandwidth into the server getting closer to the 100MB/s limit of the Fibre Channel connection.

Decisions: The Disk Drives or the Porsche?
Obviously this sort of gear isn’t targeted at the home user. It’s total overkill. But hey, it’s interesting to see what kind of technology is in use in the datacentre, that may eventually trickle down to the desktop. Storage Area Networks are (in my opinion) a great solution for dealing with huge amounts of storage. Reliable, fast, flexible and fault tolerant. It’s just a shame about the price. Personally, I’d take the Porsche. ;)

BlueSmurf (email / website)

---

Other Recent Reviews:

21-Cooler SocketA Roundup

3-Way AMD760 DDR Motherboard Roundup

Dual Palomino! On Tyan Thunder K7