The Bottom Line Is That There's No Faster Way
The results were very impressive: Up to 1.1 GB/s in RAID 0 and a maximum of about 1 GB/s in RAID 50 - those are results that have certainly raised the bar. However, it should be noted that magnetic data storage generated these impressive numbers.
In order to fully display this performance capacity, however, blocks as large as possible are required. This ensures that distributing the load across up to 32 drives runs smoothly in the first place. Using 64 kB, for example, makes almost no dent, as the 250 MB/s generated here (depending on the configuration) can be realized even with six (or more) drives.
Examining I/O performance is also interesting because, in the server environment, it is even more important than maximum transfer rates. In RAID 0, the test system generated about 2,000 I/O operations per second during the file server, database and workstation tests. To put this into perspective, individual drives notch about 150 I/O operations per second. Only during the Web server benchmark tests were the RAID 0's and RAID 50's performances close, as write operations are relatively rare.
Performance - especially in the I/O realm - will undoubtedly improve down the road, thanks to the use of more expensive hardware based on Ultra320 components. But that won't happen without massive effort, because for now, at least, no other RAID architecture can handle spanning.
