Yes you have it backwards, it is the outer part of the disk you want to use. If you had the first partition on the outer most track you might fit 5 times more data than the inner tracks so the head would never have to move to read all of the data on that one track.
A faster disk *can* be better, but it isn't always worth the cost. A 15K SAS drive is going to be fast, but you are more likely to use inner tracks in the end, which will slow things down. If you look at the links I provided you can see that using one large partition across 4 15K SAS drives is about 1.5x faster than short stroking a 1TB drive. When you start adding data to that partition you will start to lose that performance due to head movement and being smaller using the inner tracks is pretty much guaranteed. Short stroking will maintain a more consistent min/ave/max speed and will last over time. A 2TB or 3TB enterprise drive is going to have 2-3x the space on the outer tracks (either by extra platters or sectors per track), so in theory it would be even faster than the 1TB they use in that benchmark. It will also not be 3x the cost (upgrading hardware and drives) to get 1.5x performance going with SAS drives In the end if performance is really that critical doing hands on tests of each type of drive would have to be done, but it would most likely end in RAID across SLC SSDs, short stroking 600GB 15K SAS drives, or using SFF SAS drives with less dramatic short stroking. All expensive and all overkill for a lot of situations. On Fri, Apr 29, 2011 at 10:57 AM, Randall Whitman <909li...@whizman.com>wrote: > Short stroking ... thanks for the info/links. > I thought I had read once that the inner part of the disk is faster, > due to the smaller circumference for the head to traverse. > So do I have it backwards, when I put swap in the first partition? > Or is it still the case, that of the portion of the disk used > (which may less than the entirety if short-stroking), that the > innermost track of those actually in use will be fastest? > (albeit a faster disk would make more difference, as has been noted.) > /Randall > _______________________________________________ > LinuxUsers mailing list > LinuxUsers@socallinux.org > http://socallinux.org/cgi-bin/mailman/listinfo/linuxusers >
_______________________________________________ LinuxUsers mailing list LinuxUsers@socallinux.org http://socallinux.org/cgi-bin/mailman/listinfo/linuxusers
