Hardware vs Software RAID (Again) - Overclock.net

parityboy · 09-21-09

I posted this here, since it's more of an OS thing than a hardware thing. Found this article:

Quote:

Dot Hill's notion is to help volume server OEMs or ODMs (original design manufacturers) offer built-in RAID functionality without the expense of a dedicated RAID-on-chip acceleration device. They can just use a spare core in the multi-core CPU instead.

Quote:

VRA enables host-based RAID 0, 1, 5, 10 and 50 directly on motherboard SATA I/O ports, but not RAID 6. The same software stack, with additional features enabled, can also be used for higher end SAS disk-based systems that employ a built-in or PCIe host adapter-based SAS/SATA chipset, enabling vendors to maintain compatibility and a consistent look and feel across the entire product line.

Source

Obviously the proof of the pudding is in the eating, but my first thought was, "Linux and Solaris can do this already".

I can understand the argument for software RAID, especially with file systems like ZFS where data consistency checking is improved, and I can see the argument for using a spare core on a multi-core file server which would otherwise be idle.

I wonder though, since the CPU will have to do the parity calculation for every byte of data written to a parity-based array, wouldn't there be a lot of back-and-forth across the PCIe lanes? Obviously, this is much more of an issue with PCI.

Opinions?

		Overclock.net - Overclocking.net > Software, Programming and Coding > Operating Systems > Linux, Unix
Hardware vs Software RAID (Again)

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)