On Tue, 18 Jul 2006, Greg Freemyer wrote:
On 7/18/06, Jon Nelson <jnelson-suse@jamponi.net> wrote:
As an example, in a situation involving fast ethernet (not gig-e) and some mid-grade machines (just shy of 1GHz and basic consumer-grade IDE disks) my local disks top out around 11.45 MB/s (hdparm -t), averaging 8.9 MB/s (dd if=/dev/hda of=/dev/null bs=4k count=10000). Contrast that with a recent test I performed using ATA over Ethernet, where the disks were a raid5 on a similar machine but faster disks (about 20MB/s local read speed on the backing store for the ata-over-ethernet block device) than, my local disks. (Granted, not the best test components or the best
- the exact same client was able to sustain 11MB/s and with less load, and not-very-good ethernet cards. By this quick test, and measuring "bandwidth", ata-over-ethernet was just as fast as, and sometimes faster
test, but it serves to illustrate.) With gig-e one could easily exceed my local disk speed on a regular basis.
Question: what might be the best way to calculate "latency" in a way that could be used to compare (for example) ATA-Over-Ethernet to a Local Disk (both are block devices to the client).
Jon,
Your local speeds seem slow (i.e. unoptimized). I spent a lot of time a few years ago doing speed tests and getting good disk controllers in my machines.
Like I said, these are old drives and not very fast machines. A Duron 750 with 2 3.2G (5400rpm, 6 years old) drives in it does not qualify as "fast". Both drives are old enough that they don't support SMART and smartctl (even with the hoop-jumping options turned on) can't tell me much about them. -- Carpe diem - Seize the day. Carp in denim - There's a fish in my pants! Jon Nelson <jnelson-suse@jamponi.net> -- Check the headers for your unsubscription address For additional commands send e-mail to suse-linux-e-help@suse.com Also check the archives at http://lists.suse.com Please read the FAQs: suse-linux-e-faq@suse.com