View previous topic :: View next topic |
Author |
Message |
confusion Tux's lil' helper
Joined: 24 Mar 2004 Posts: 132
|
Posted: Sat Jun 04, 2005 11:20 pm Post subject: Response difference between athlon 2600 and sparc system |
|
|
I'm thinking about ditching my athlon 2600 (2ghz) system in favour of either an ultra60 configuration or perhaps a blade 1000. I'd like to know what kind of performance difference i would experience with this setup. I know it must be slower than my 2600 configuration but how much so? Would the likes of KDE and X be much less responsive, and would load times be a lot worse? I wish there was a better way of comparing cross-architecture performance than word of mouth.
Thanks in advance,
John |
|
Back to top |
|
|
labrador Guru
Joined: 04 Oct 2003 Posts: 316
|
Posted: Sun Jun 05, 2005 3:06 am Post subject: ultra60 is much slower, tuned to server role |
|
|
I'm not sure on the disk speed and CPU speed on the Blade systems, but
a Ultra 60 based system is much slower than a AMD Athlon 2600.
I have an Athlon XP 2200 and Ultra 60 300Mhz. The Ultra
is not able to run fully featured DRM from the Elite 3D video.
In addition, the disk speed is much slower on Ultra SCSI than
on modern IDE UDMA systems.
One speed test that probably tells alot is how long emerge sync
takes on both. An AMD Althon 2200 x86 system is able
to complete it in a few minutes - perhaps up to 5 minutes maximum.
An Ultra 60 system can take up to 15 minutes to complete
emerge sync (even though the Ultra box is at a University with
much better bandwidth, while the Athlon machine is on DSL connection).
I'm not sure why you are thinking ultra 60 or blade is better than
the Athlon - perhaps you are looking at the 64 bit issue.
Well, that is only available for the kernel, not applications, on Sun platform.
The thing that drags down performance on the Suns is
disk I/O speeds and CPU Mhz. On the other hand Sun machines
do "survive" being driven by multi-user demands better than x86, so
for a server role it isn't bad. |
|
Back to top |
|
|
confusion Tux's lil' helper
Joined: 24 Mar 2004 Posts: 132
|
Posted: Sun Jun 05, 2005 12:27 pm Post subject: |
|
|
Yeah I kinda figured as much, worth a shot just to be sure though. I think i might just replace my ultra5 server instead. I've heard of people putting gigabit cards and IDE controllers in to ultra's before. If i got hold of an ultra 30 or 60 do you think it would be possible to run it as a fast file server? The ultra5 has been really reliable so far, i've developped a taste for sun hardware lol.
--John |
|
Back to top |
|
|
labrador Guru
Joined: 04 Oct 2003 Posts: 316
|
Posted: Sun Jun 05, 2005 6:14 pm Post subject: it's all relative to what you want to do... |
|
|
I don't know what you mean by fast file server.
If you mean able to deliver maximum file transfers for just you or a couple of
others, then you probably want something with higher transfer speed drives.
I've not heard of anyone putting IDE drives in the Ultra 30/60/80.
Housing them and possibly supplying power to them, are the main obstacles,
and you probably would not be able to boot from them.
I don't have any real figures, but I think if you had a SCSI based system
and an IDE based system, and both were getting hit on by many file
service visitors, the SCSI based system would be able to regulate the
traffic better. If you were to make an analogy with an airport,
the advantage with the Sun Ultra 60 would be in the better air traffic
control. So it isn't so much a question of how fast planes go,
but how many planes you can handle per minute. In file servers,
if you can't respond to the requests when there is a high number of
visitors, then it doesn't matter what your ethernet, CPU speed and
MB/s transfer rate are.
Perhaps I can leave you with an example of this. At the University where
I work, there is an Ultra 1 with an external storage array, which
provides all of the email POP services for roughly 4000 students.
The machine is very busy. I wrote a perl script that would provide
a text report of accounts that were over quota. On an unloaded
Ultra 1, using the same test data, the perl script ran in 2 seconds.
On the production Ultra 1, the script took 30 minutes against the
same data. It is that punished, and yet people do not get
a timeout when retreiving their mail by POP! I would guess that
an x86 with fast IDE drives would not perform as well,
due to the lack of SCSI features.
Perhaps this article on a debian mailing list can best describe why
and where SCSI has advantages:
http://lists.debian.org/debian-user/2004/01/msg02091.html
For single user or something closer to that, here are some hdparm ratings
to compare from my systems:
Ultra 60 (300Mhz) with 10K RPM Ultra SCSI:
Code: | # hdparm -tT /dev/sdb
/dev/sdb:
Timing O_DIRECT cached reads: 940 MB in 2.00 seconds = 470.00 MB/sec
Timing O_DIRECT disk reads: 82 MB in 3.10 seconds = 26.45 MB/sec
# hdparm -tT /dev/sda
/dev/sda:
Timing O_DIRECT cached reads: 936 MB in 2.00 seconds = 468.00 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.04 seconds = 21.05 MB/sec
|
PIII 650 (dual proc) with UDMA 133 controller and Seagate IDE drive:
Code: | # hdparm -tT /dev/hdg
/dev/hdg:
Timing cached reads: 572 MB in 1.99 seconds = 287.44 MB/sec
Timing buffered disk reads: 168 MB in 3.01 seconds = 55.81 MB/sec |
If you want to go with the Ultra and SCSI, get 10K RPM disks. The 7200 RPMs
show about 12 MB/sec |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|