On 26 Feb, 22:23, Ilya Zakharevich <nospam-ab...@ilyaz.org> wrote:
> [A complimentary Cc of this posting was sent to
> Martin Pearce
> <martin.pea...@ieee.org>], who wrote in article <de30c7ce-2e9c-4928-9c28-dc0c4e7d9...@o77g2000hsf.googlegroups.com>:
>
> > > I guess my tinkering may have paid off. *Initially the device seems to
> > > be *significantly* faster (3-5x) then the previous swap mechanism (a
> > > single slice on a 146GB 10K SCSI drive).
>
> One should reasonably expect about 50MB/sec throughput there, right?
> (If it is on par with contemporary PATA drives...)
>
> > Actual performance appears somewhat disappointing in comparison with
> > the original posting.
> > (maybe it's in the configuration of the device.).
>
> You understand that USB sticks can be as low as 2MB/sec (as my 1GB one is)?
>
> Yours is a little bit newer, but expecting much more than about
> 6MB/sec should be accompanied by a *specific* search for high-speed models..
>
> Hope this helps,
> Ilya

Yes that would be consistent with my original view.

This was a bit of an experiment, as this particular model of USB flash
memory is advertised by SanDisk (in wintel circles) as being
ReadyBoost compatible. This technology I understand is supposed to
include device access-time enhancements for Paging (rather than pure
MB/s transfer speeds).

As this is a generic device, with no additional drivers required for
Solaris, I was interested in seeing if the technology could be levered
on to non-wintel platforms. I have yet to check thoroughly, but I
think there might be some activity in the Linux arena with this kit.

(as mentioned I'm using USB-2 on then SB1000)


Regards Martin