System setup and notes
We're nearly at the testing stage now, but first, system setup and notes.
| Component | Details |
|---|---|
| Processor | AMD Athlon 64 3200+ (Socket 939, Winchester, 2.0GHz) |
| Memory | 2x 512MiB Corsair ValueSelect PC3200 |
| Graphics | S3 Virge PCI |
| Motherboard | ASUS A8N SLI Deluxe - BIOS 1017-004 |
| Hard drives | Seagate ST380817AS 80GB SATA 1.5Gbps 4x Seagate ST3750640AS 750GB SATA 3Gbps |
| Storage controllers | NVIDIA SATA - nForce 6.70 drivers LSI MegaRAID 8408E - v1.20 driver |
| Operating System | Windows XP Professional 32-bit SP2 |
Notes
Feel free to laugh at the S3 Virge PCI. There wasn't a spare PCIe graphics card at this reviewer's disposal at the time of testing. That said, we're not silly enough to test a RAID array with 3DMark 2006. The graphics subsystem isn't what's on test, so it doesn't matter. With no PCIe graphics card, we popped the MegaRAID into a PEG slot. However, had we a PCIe card we could have run both; with the board in SLI mode both PEG slots get x8 PCIe bandwidth; still enough for the MegaRAID.
Other observations we might make on the system specification include the CPU. A server running this card may have a multi-core or multi-CPU setup. There may also be more (and possibly DDR2) RAM.
Finally, the operating system in use is subject to a 2TB disk size limitation. Our four 750GB disk RAID-5 array weighed in slightly above this (as did RAID-0, obviously). This can't be fixed by only formatting part of the disk... disk manager can't see it because it's too big. Luckily, the RAID management tools allowed us to reduce the size of the virtual disk presented to the OS. We trimmed it down to just below 2TB where necessary, leaving a portion of the disks unused, but all of them still part of the array, meaning performance shouldn't be affected. 64-bit versions of XP and Windows Server 2003 are not limited to 2TB volumes.
Testing software
IOMeter 2004.07.30
IOZone 3.263
HD Tach RW 3.0.1.0
For IOZone, we ran our usual read/write tests with file sizes up to 128MB and record sizes up to 16MB on a single, NTFS formatted-partition on the drives/arrays.
HD Tach was run on unformatted disks, allowing us to undertake write tests to verify our other write-test results. However, we'll only be graphing CPU and burst results.
Finally, here's our testing regime for IOZone, also applied to unformatted disks:
| Option/Test | Configuration |
|---|---|
| Outstanding I/Os | 10 |
| Individual test run time | 30 seconds |
| Read test access spec | 1MB transfers 100% sequential 100% read |
| Write test access spec | 1MB transfers 100% sequential 100% write |
| General usage access spec | 64KB transfers 50% sequential, 50% random 33% write, 67% read |
Disk configurations
We used the following disk configurations for our testing, to test out the RAID card's capabilities as well as its performance.
- MegaRAID level 5, in online, degraded and rebuilding states, default settings
- MegaRAID level 5 alternative config. with write-back caching and read-ahead enabled
- MegaRAID level 0
- Single 750GB Seagate SATA disk
- NV SATA RAID-0 using 2x 250GB Seagate SATA disks
We fiddled with a few performance settings to come up with the alternative configuration for the RAID-5 array, to see if it would make any difference.
While not tested on exactly the same system or with the same disks, we provided the NV RAID-0 results as some form of comparison against what is essentially pretty cheap (and not failsafe) RAID.
