I've had good luck with the 8port 3ware 7850 model. I benchmarked it against software raid under a fully updated dual-cpu system running Redhat 7.2 w/ ext3 a while back using the iozone software package. For RAID-0, the 3ware controller significantly outperformed the md driver for read performance. It looks like the 3ware drivers spread reads across disks to maximize bandwidth. The md driver performed almost identically to reading from a single disk. For all other operations, the 3ware driver was slightly slower than the md driver. I'd attribute this to the additional layer of SCSI emulation overhead between the IDE drives and the OS. The other thing of note was that 3ware driver performed sub-optimally on all tests when reading or writing files that were exactly 16MB in size. The iozone graphs should show this clearly if you run the benchmark yourself. I have no idea why this would be true, and have not pursued it with 3ware. The 3ware cards have been stable after updating the default kernel driver to the driver published on 3ware's web site. They have been a lot more stable than similar complex IDE drive configurations using promise controllers. On Fri, 2002-08-30 at 12:52, Ed W wrote: > Slightly OT, but this looks like a good place to find a lot of disk > controller experience. Since I need to swap out some dodgy hardware anyway, > I am thinking about upgrading to one of the 4 port 3ware IDE RAID cards. It > will only be in a 2 disk mirror config, however, I am hoping that it will > give a nice speed boost > > Any thoughts on compatibility with dual cpu machines? Any one have any > anecdotes on stability or compatability? Are there any pros or cons of > going up or down the range, ie older 32 bit PCI is much less money, driver > compatability, etc > > Also, any special config of this card needed for best performance under > ext3? (probably not?) > > Thanks, > > Ed W > > > > > > > _______________________________________________ > > Ext3-users@redhat.com > https://listman.redhat.com/mailman/listinfo/ext3-users
