I just got a new PC which, after many MANY woes, just started to run OK. In case you are interested, I got an Asus P5B-VM DO miniATX motherboard whose onboard soundcard was crackling, and installed my Audigy on the one precious spare PCI slot, and it was popping. I finally relented and got a Gigabyte P35C-DS3R which is just awesome (not as many features as the Asus, but it works great).
I wanted a RAID5 array, so I got 3 500 GB disks and slapped them on. One disk was DOA so I sent it to be replaced, leaving me with 2 disks and nothing to do with them. Well, there was one thing to do, and that was: stripe them. I did so, and having a 10k RPM hard disk for the OS itself, I decided to benchmark them to solve the age old question. I used Everest’s hard disk benchmark feature, anticipating that the RAID array would be a bit better at throughput and the 10k RPM Raptor a bit better in latency, but I didn’t know exactly by how much. Well, here are my results:
RAID Stripe (7200 RPM Western Digital 500 GB disks)
{width:20em}. Linear Read (Begin) 166.7 MB/s Linear Read (Middle) 135.2 MB/s Linear Read (End) 78.4 MB/s Random Read 76.6 MB/s Buffered Read 380.0 MB/s Average Read Access 12.92 ms
10k RPM Western Digital Raptor 36 GB
{width:20em}. Linear Read (Begin) 84.8 MB/s Linear Read (Middle) 76.9 MB/s Linear Read (End) 55.9 MB/s Random Read 66.5 MB/s Buffered Read 121.8 MB/s Average Read Access 7.78 ms
I don’t have my RAID 5 array yet, but I would expect that these results would be identical, and there would be a small delay in writing. Based on these results, I’d take the RAID array any day (especially because the size of the array is 1 TB versus 26 GB of the 10k). The array is almost twice as fast as the 10k disk, with not really that much of a latency difference (I guess this makes me a hypocrite, because I call 135.2 almost double of 77 MB/s while calling 12.92 a negligible difference over 7.78 ms, but the thing is that 5 ms isn’t that big of a deal, while 77 MB/s is). If you’re a speed freak, you could stripe two Raptors and do even better, but that’s too expensive for not enough gain for me.
If you have any doubts or wish to contest my results (which I encourage), please post comments here.
Edit: The RAID5 results are back from the lab:
RAID 5 (3 7200 RPM Western Digital 500 GB disks)
{width:20em}. Linear Read (Begin) 165.0 MB/s Linear Read (Middle) 137.7 MB/s Linear Read (End) 77.1 MB/s Random Read 119.6 MB/s Buffered Read 1953.5 MB/s (yes, 2 GB/s) Average Read Access 13.26 ms
Here’s another one, for perspective:
Plain 250 GB Seagate 7200 HDD
{width:20em}. Linear Read (Begin) 67.8 MB/s Linear Read (Middle) 60.6 MB/s Linear Read (End) 35.1 MB/s Random Read 64.4 MB/s Buffered Read 87.4 MB/s (yes, 2 GB/s) Average Read Access 15.16 ms