Re: [COMP] Fw:
- From: John Madden <weez@xxxxxxxxxxxxxx>
- To: computers@xxxxxxxxxxxxxx
- Date: Fri, 25 Feb 2000 17:41:25 -0500
> Can someone do a similar test under Linux and post the results? My Linux
> computer is very slow and wouldn't give me a realistic measurement. I'm
> rather curious.
This turned out to be an interesting experiment. I tried it on 3
machines, all Linux.
Disk to disk, different controllers: 7 seconds, or 1.43 megs/sec
Disk to disk, same controller: 3 seconds, or 3.33 megs/sec
Disk to disk, same partition: 4 seconds, or 2.5 megs/sec
System specs: Linux 2.2.6, Pentium 75, 32 megs Ram. Running
apache/mysql/user accounts/RC5/other misc departmental stuff, and
**16-bit Disk access**. Turn on 32-bit access, and I get:
Disk to disk, different controllers: 3 seconds, or 3.33 megs/sec
Disk to disk, same controller, same partition: ~3.5 seconds, or 2.89
megs/sec
Disk to disk, different controllers: 3 seconds, 3.33 megs/sec
The numbers here may be a bit off. I would've expected that turning on
32-bit access would've improved things a lot more than it did. I did
this remotely (ssh is great) on a machine at work, so network lag may
have delayed my noticing the file copy's completion. Also, this is a
production server- things may have been going on that were eating up cpu
cycles.
On Avenir web, it took 12 seconds to copy the 10 meg (exactly) file from
one point on the partition to another (sorry, only one hard drive) with
16-bit disks, and 10 seconds with 32-bit access. Avenir Web's a 486
with 20 megs ram, running Linux 2.2.2.
On my machine (again, only 1 hard drive), it took 6 seconds at 16-bit
(that's 1.67 megs/s), and 4 seconds (2.5 megs/s) at 32-bit. My
machine's a P233, 96megs ram, running Linux 2.2.13.
All hard drives in this little benchmark are 5400 rpm IDE drives.
There's a mix of Quantum, IBM, and Seagate drives.
The 10-meg file, in case you're wondering, was created with
dd if=/dev/hda of=temp count=10000 bs=1k
Linux obviously has some performance advantages, but again, the numbers
aren't exact. I'm just watching wmCalClock, and counting the number of
seconds before the file stops copying. :) A 10-meg file is pretty
large. Smaller files (ones around a meg, for example) will copy much
more quickly, because the OS has a better chance of caching the data.
Hard drives nowadays also have nice large caches, meaning the data
doesn't actually have to be written to disk right away. The Linux
kernel will also cache things a lot better than Win9x. There are a lot
of factors to consider here, so take the "benchmarks" with a 10-meg
grain of salt. :)
John
--
# John Madden weez@xxxxxxxxxxxxxx ICQ: 2EB9EA
# Sys-Admin / Webmaster, Avenir Web: http://avenir.dhs.org
# LANdb: Network Admin Database - http://avenir.dhs.org/landb/
# "A kernel compile a day keeps the blue screens away."
========================================
Avenir Web's Computers Mailing List
List Modes, Subscription, and General Info:
Go to http://avenir.dhs.org/mailing.html
List Archives: http://avenir.dhs.org/archives/
Administrative Contact: weez@xxxxxxxxxxxxxx
Get computer help: http://avenir.dhs.org
========================================
- References:
- RE: [COMP] Fw:
- From: Michael V. Franklin
Other related posts:
- » Re: [COMP] Fw:
- » Re: [COMP] Fw:
- » Re: [COMP] Fw:
- » RE: [COMP] Fw:
- » RE: [COMP] Fw:
- » Re: [COMP] Fw:
- RE: [COMP] Fw:
- From: Michael V. Franklin