On Fri, Jun 26, 2009 at 11:55 PM, Matt Kivela<[email protected]> wrote: > Hi Guys: > > Hopefully someone can point me in the right direction since I can't quite > find it with searching around... > > I'm doing some benchmarking with open source software using solid state > drives. We know from non-OS dependent tests that these drives with 64k > blocks can do about 220MB/s sequential reads, but with 4k blocks only about > 90MB/s. > > Linux won't accept block sizes > page size, and on x86 architectures Linux is > limited to 4k pages. Thus we have a wall of about 90MB/s it doesn't seem > like it's possible to pass. I'm not sure if this is an Intel x86 limitation, > or a Linux limitation, or simply a combination of the two -- since Linux on > other platforms can have larger page sizes.
Which target on Linux? I use the iSCSI Enterprise Target on Linux and with blockio I can do variable block sizes up to the controller's max segment size (640K on my controller), of course with blockio it does no system memory caching, so you will need a disk controller with on board cache for it to perform under load. -Ross _______________________________________________ storage-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/storage-discuss
