|
Eight-Drive SATA II RAID 0 on a PCIe G5 PowerMac I just purchased a new PowerMac G5 Quad Core for HD and SD video editing in Final Cut Pro, along with a BlackMagic Design Multibridge Extreme. Please Apple and BlackMagic -- no more products with marketing buzzwords like "Pro" or "Extreme". I beg you. But I digress. HD video editing and compositing require tons of drive space and speed, and there are many good options out there, including Apple's XServe RAID as well as the Huge and Medea products, which are all quite good and also quite expensive. I had a very good experience with the last SATA RAID I built, so I decided to give it another go. My requirements were a PCIe-based system, that would allow for expandability as well as maintain speeds of 300 MB/sec or more through 90% of the drive space. After looking at my options, I decided on the Sonnet E4P, which is a "port multiplier aware" card. Port multipliers allow up to 5 drives to operate off of a single SATA II channel. The E4P has 4 channels, so it has the capability of attaching 20 drives total. Fill those with 500 GB drives and you max out at 10 TB. The downside is that all the drives on a multiplier have to share the 225 MB/sec max (according to Sonnet and other sources) in each channel. On the other hand, the speeds are more likely to be consistent through the entire capacity of the drives. I chose a Burly 8-drive case with 2 port multipliers which I got from MacGurus. This will leave over 2 channels on the card for future expansion. For drives I got 8 Seagate 7200.9 SATA II 300 GB drives (with 16 MB caches) for a total formatted size of 2.2 TB for just under $2000 for the card, drives and case with tax and shipping. After putting it all together, which took about half an hour, I hooked it up to the new G5 and built the RAID 0 set using Disk Utility. Using a folder which had about 200 GB of media in it, I started testing how drive speeds were effected as they became full. The tests were done with BlackMagic's Disk Speed Test, which writes and reads actual video frames from the Multibridge to test the disks. Each result listed is the average of 10 separate trials. You can download a PDF of the full results or the Excel Worksheet. Results The first result of interest is that empty, the drive speeds are quite adequate for all flavors of SD and HD, averaging almost 380 MB/sec on both reads and writes. This was about what I expected, with a maximum theoretically of about 450 MB/sec (2x225).
What was more surprising, however, was that unlike the SATA RAIDs I've seen without port multipliers, read speeds actually INCREASED (peaking at over 400 MB/sec) through about 60% of the drive capacity before they started decreasing. Write speed remained very stable through 75% of the drive capacity before tailing off.
Another thing I wanted to look at was consistency, both over the first 75% of the drive capacity, as well as among individual trials. Using the standard deviation measure, we can see that read speeds are more erratic than writes, which are rock solid consistent both within each set of trials and through the first 75% of drive capacity. Even so, the variance in reads is well within the acceptable range; only 2 trials out of 80 were under 370 MB/sec in the first 75% of capacity. Drive speeds did drop considerably at the very end of capacity but this is to be expected and speeds are still within acceptable ranges for most uses even to the last bytes. Also note that the variance in the read trials dropped considerably in the final 25% of drive capacity as the results became more consistent.
The standard deviation for all read samples in the first 75% of drive capacity was 11.61. The same for writes was only 3.19. Conclusion This RAID is an excellent choice for video or film work. It is fast, consistent, large, and expandable. It is usable for all flavors of HD and SD through the first 90% of drive capacity, and good for everything up to and including 720p all the way through to the last byte. |
