FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Debian > Debian User

 
 
LinkBack Thread Tools
 
Old 12-12-2008, 03:33 AM
"Mag Gam"
 
Default server upgrade question

At my university we have 10 servers. Each server has 8 cores with 32
GIG of memory running Debian 4.0. We have to give these servers to a
different department, and our Dean would like to consiladate 10
servers into 5 servers. The new server will have 16 cores with 64 GIG
of memory. Basically a 2:1 type of deal.

Since we are doing a 2:1, should we expect 2:1 performance? For
instance, most of our applications are heavy compute and memory
intensive applications. Would they run at the same speed, better, or
worse with this new setup? My guess is that same?

Oh, yeah will be running 4.0 :-)

TIA


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-12-2008, 05:47 AM
lee
 
Default server upgrade question

On Thu, Dec 11, 2008 at 11:33:09PM -0500, Mag Gam wrote:
> At my university we have 10 servers. Each server has 8 cores with 32
> GIG of memory running Debian 4.0. We have to give these servers to a
> different department, and our Dean would like to consiladate 10
> servers into 5 servers. The new server will have 16 cores with 64 GIG
> of memory. Basically a 2:1 type of deal.
>
> Since we are doing a 2:1, should we expect 2:1 performance? For
> instance, most of our applications are heavy compute and memory
> intensive applications. Would they run at the same speed, better, or
> worse with this new setup? My guess is that same?
>
> Oh, yeah will be running 4.0 :-)

Huh? 10 servers * 8 cores * 32GB makes 80 cores and 320GB. 5 servers *
16 cores * 64GB makes 80 cores and 320GB. Even assuming that all the
cores operate at the same speed (which depends, for example, on how
many cores per CPU there are), you'll lose a lot of performance
because the same amount of cores and memory will have to share 1/2 the
amount of servers. That might cut your performance in half.

The other way round: Imagine you have two computers with one CPU, one
core each. You have 16GB of memory in each. Now you replace the two
computers with one that has one dual core CPU and 32GB of memory. It
will be slower because instead of having "dedicated" resources for
each core, you now have several cores sharing the resources. If you
replace them with a computer that has two single core CPUs on the
board, it will also be slower because two CPUs share the resources.


Other than that, the performance can better or worse, there are too
many other factors that influence it. You'd have to do some benchmarks
to tell.


--
"Don't let them, daddy. Don't let the stars run down."
http://adin.dyndns.org/adin/TheLastQ.htm


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-12-2008, 01:00 PM
Napoleon
 
Default server upgrade question

Mag Gam wrote:

At my university we have 10 servers. Each server has 8 cores with 32
GIG of memory running Debian 4.0. We have to give these servers to a
different department, and our Dean would like to consiladate 10
servers into 5 servers. The new server will have 16 cores with 64 GIG
of memory. Basically a 2:1 type of deal.

Since we are doing a 2:1, should we expect 2:1 performance? For
instance, most of our applications are heavy compute and memory
intensive applications. Would they run at the same speed, better, or
worse with this new setup? My guess is that same?

Oh, yeah will be running 4.0 :-)

TIA




Too many variables to tell, but it won't be 2:1. That won't be possible
due to resource contention.


If you aren't using your existing cores effectively, you might not see
any performance gain on a server-for-server basis. But for typical
systems, you might expect a 25-50% increase over an existing system.


That is, of course, assuming all other things (CPU speed, hardware cache
sizes, disk speed, etc.) remain the same. Changing those parameters
makes things more complicated.



--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-12-2008, 01:08 PM
"Douglas A. Tutty"
 
Default server upgrade question

On Thu, Dec 11, 2008 at 11:33:09PM -0500, Mag Gam wrote:
> At my university we have 10 servers. Each server has 8 cores with 32
> GIG of memory running Debian 4.0. We have to give these servers to a
> different department, and our Dean would like to consiladate 10
> servers into 5 servers. The new server will have 16 cores with 64 GIG
> of memory. Basically a 2:1 type of deal.
>
> Since we are doing a 2:1, should we expect 2:1 performance? For
> instance, most of our applications are heavy compute and memory
> intensive applications. Would they run at the same speed, better, or
> worse with this new setup? My guess is that same?
>

Will the new servers have double the number of busses, double the hard
drive throughput, double the memory bandwidth?

Doug.


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-15-2008, 03:54 PM
"Stackpole, Chris"
 
Default server upgrade question

> From: Mag Gam [mailto:magawake@gmail.com]
> Sent: Thursday, December 11, 2008 10:33 PM
> Subject: server upgrade question
>
> At my university we have 10 servers. Each server has 8 cores with 32
> GIG of memory running Debian 4.0. We have to give these servers to a
> different department, and our Dean would like to consiladate 10
> servers into 5 servers. The new server will have 16 cores with 64 GIG
> of memory. Basically a 2:1 type of deal.
>
> Since we are doing a 2:1, should we expect 2:1 performance? For
> instance, most of our applications are heavy compute and memory
> intensive applications. Would they run at the same speed, better, or
> worse with this new setup? My guess is that same?
>
> Oh, yeah will be running 4.0 :-)
>
> TIA

It really depends. If your applications talk to each other a lot,
chances are you will see an increase in speed when they don't have to go
out to the LAN.

You say your applications are Memory intensive and this will probably
have the biggest impact. If the application is greedy and just uses all
of the memory it can get its hands on, then you will probably see a
decrease in performance as the instances of the applications will fight.
If the application uses a set amount of memory (eg 2GB) and just
constantly read/writes to that portion, then you probably won't notice a
difference.

When dealing with applications across multiple systems/cores it is very
important to determine exactly what your overhead and constraints are
first before trying to "upgrade" the system.

A good example is a cluster I worked on a few years ago. It started out
as five P3 500Mhz boxes. When we upgraded to fifteen Athlon 1.2Ghz
systems, our application slowed to a crawl. The cross-talk on the LAN
connection was killing the app. We temporarily configured just 2 of the
Athlon systems and got better performance then either of the previous
configurations while we recoded the app to better deal with scalability.

If you can identify where the applications bottlenecks and strengths
are, you will be in a much better position to know how hardware upgrades
will affect your results.

Have fun!
~S~


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 

Thread Tools




All times are GMT. The time now is 10:22 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org