Because we are looking at these SATA drives from an enterprise deployment perspective, I tested their performance accordingly, using Iometer 2003.2.15 to simulate typical server loads and recording the number of I/O operations per second that each drive was able to sustain. A second Iometer script was used to measure the maximum transfer rate. I ran each script for two minutes, excluding the first six seconds of activity and recording the final average results.
The test bed was an HP ProLiant ML350, with two Pentium 2.2 GHz processors and Windows 2000 Advanced Server SP3. The tests were run on single, un-partitioned drives connected to an Adaptec 1210SA controller.
The 73GB Fujitsu MAS3735NP doesn’t really belong to the SATA group, but I applied to this drive the same testing criteria to provide a reference point with a fast-performing SCSI 320 unit. The Fujitsu SCSI drives were attached to an Adaptec 29320-R Ultra320 controller.
The server load simulation test involved data transfers varying from 512 bytes to 64KB, discretely scaled according to a commonly-accepted pattern for file server load. For each data transfer size, the script created a composite mix of random disk accesses, splitting I/O operation between reads and writes with an 8/2 ratio.
To increase drive stress factor, I started two concurrent processes (Iometer workers) for each script. To simulate light, moderate, and heavy server loads, I ran the same script three times against each drive, setting the number of outstanding I/Os to one, then to 16, and finally to 64.
To measure the maximum transfer rate in MBps (megabytes per second) that each drive can deliver in read or write mode, I ran separate tests, involving two Iometer scripts (one for read, the second for write operations), each generating 64KB-sized data transfers.
Reliability scores consider the vendors’ shock resistance and error rate published specs for each drive, and the warranty terms offered.