sean | 
06 27th, 2009| 
A quick rundown of dbench results and hardware used. Results seem really bad for RAID systems, and it’s funny how a 5+ year old IDE RAID 5 card with 16/32MB RAM (I forget) can get ~double the performance of an $800 enterprise-level SAS / SATA controller with 512MB DDR2 RAM.
And so these should be taken with a large grain of salt.
T1.) Throughput 93.9068 MB/sec 200 procs
mercury (2x Xeon 5120 @ 1.86Ghz (4 cores total, 4096k L2 cache / 2 cores), 3GB RAM DDR2, CentOS 5.3)
/home# single 80GB Hitachi SATA drive
2.a) Throughput 19.9959 MB/sec 200 procs
atlas (Pentium D @ 3.0Ghz (2 cores total, 2048k L2 cache), 4GB RAM DDR2, Debian Lenny)
/home/likwid# LSI Logic 8480E RAID 5 w/ 3x 1TB WD Green series SATA drives
2.b)Throughput 15.4584 MB/sec 200 procs
atlas (Pentium D @ 3.0Ghz (2 cores total, 2048k L2 cache), 4GB RAM DDR2, Debian Lenny)
/# Linux Software RAID 2x 80GB SATA drives
3.a) Throughput 11.9199 MB/sec 200 procs
nep (Pentium 4 1.8 (256k L2 cache), 1024mb PC133 SDRAM)
/home/spetrow# 3ware Escalade 6200 RAID 1 w/ 2x WD 40GB IDE drives
3.b) Throughput 14.7577 MB/sec 200 procs
nep (Pentium 4 1.8 (256k L2 cache), 1024mb PC133 SDRAM)
/backup# single WD 320GB IDE drive
4.a) Throughput 40.4807 MB/sec 200 procs
z (Pentium 4 2.0 (512k L2 cache), 1280mb PC133 SDRAM)
/home/likwid# 3ware Escalade 7506-4LP RAID 5 w/ 3x WD 500GB IDE drives
4.b) Throughput 49.3065 MB/sec 200 procs
z (Pentium 4 2.0 (512k L2 cache), 1280mb PC133 SDRAM)
/# 3ware Escalade 6200 RAID 1 w/ 2x WD 40GB IDE drives