Disk throughput Questions

From: <booysems%telkom09_at_qtts-nfs-2303.telkom.co.za>
Date: 03 Mar 95 14:47:03 GMT

Hi,

I have the following configuration:

Alphaserver 2100 wirh 512MB memory

7 x 1GB rz26l disks configured in a RAID with one hot standby disk. That
means I have 5GB usable space in the RAID. The RAID controller is a 3
port controller on the EISA bus.

1 x 2GB rz28 (System disk), 3 x 1GB (Swap) on a PCI SCSI controller.


I want to test the throughput on all these disks (raw and filesystem).
I used the following commands and received these results:

1.) RAW RAID:

    time dd if=/dev/mem of=/dev/re0c bs=64k ---> 821K/s throughput


2.) RAW standalone disk:

     time dd if=/dev/mem of=/dev/rz13c bs=64k ---> 578K/s throughput


3.) FIlesystem on RAID:

     time dd if=/dev/mem of=/dev/raidfs/mem bs=64k ---> 2.7MB/s
throughput


4.) Filesystem on standalone disk:

    time dd if=/dev/mem of=/dev/sfs/mem bs=64k ---> 1.9MB/s thoughput


Questions:
a) Why is the Filesystem tests better that the raw devices?
b) Shouldn't I get at least a throughput of 3MB/s on the raw devices?
c) Seeing that the PCI bus is a lot faster that EISA, why is the RAID
faster than the standalone disk?
d) Are the tests that I have performed valid at all? Is there another
more valid method to use?



Any replies will be appreciated!



Thanx.
Marius Booysen
Email: booysems_at_telkom09.telkom.co.za
Received on Fri Mar 03 1995 - 08:30:23 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:45 NZDT