On Sat, Jan 19, 2013 at 03:55:17PM -0800, Linda Walsh wrote: > > > Dave Chinner wrote: > >On Sat, Jan 19, 2013 at 11:47:44AM -0800, Linda Walsh wrote: > >>Given that XFS relies on the page size for it's maximum block size, > >>does switching to 1GB pages for kernel pagetables allow for larger > >>page-size use in XFS -- and therefore, larger blocksize? > > > >No, it doesn't. The page cache can't use large pages so there's no > >path to allowing XFS to use larger blocks with them. > > > >As it is, you'd need a system that supports 16k or 64k page sizes > >for this to be useful, otherwise you just waste a *lot* of memory... > > > >>Is it something that might be in the not too distant future? ;-) > > > >I have no plans to do this any time soon - there's no compelling > >reason to do so. I can't speak for anyone else, though... > > --- > No prob.. I admit a 1GB page size might be a tad large for > my purposes... I actually was reading 1MB though even that would be > large...but a GB/disk sector...um...-- at least I'd be likely to > get full speed read & writes if I could fill them...(1GB/R or W > gives me about the fastest overall throughput to my disks). > > Can't test higher (i.e. >=2GB) than that due to getting > partial block read/writes (on a 64 bit machine??) Sounds like I think that the read/write interface are limited to 2GB buffer size, regardless of the platform. > my LSI controller is only 32bits...but that doesn't seem right... > Maybe it only supports consecutive read/writes up to the limit > of the memory on the controller card.. > > All that talk about RAIDs recently, got me depressed a bit > when I realize that while I can get fast speeds, type speeds in seeking > around are about 1/10-1/20th the speed...sigh. > > Might that indicate that I should go with smaller RAIDS with more > spindles? I.e. instead of 3 groups of RAID5 striped as 0, go for 4-5 groups > of RAID5 striped as a 0? Just aligning the darn things nearly takes a rocket > scientist! But then start talking about multiple spindles and optimizing > IOP's...ARG!...;-) (If it wasn't challenging, I'd find it boring...)... Somebody on the list might be able to help you with this - I don't have the time right now as I'm deep in metadata CRC changes... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs