[SUMMARY] OT: High End Servers comparison

Thanks to 
Chris Hoogendyk 
Wiest, Damian 

Chris advised to use transaction per second as a metric for comparison ! 
 ( typical database and banking benchmarks are transactions per second.)
 
Below is Damian words it was very useful for me:
==============================================================
Hi Levi,

We just did a comparison of some Sun and IBM servers for a new project.
Here's some of the typical things to look for:

Component redundancy: power supplies, system boards, etc.
Number of CPUs and type supported
Amount of core memory supported
Type and number of disks supported: IDE/SCSI/fiber channel RAID support
Network interfaces PCI slots, including type/speed of slots Remote
monitoring capabilities Supported operating systems Miscellaneous I/O ports:
USB, SCSI, audio, etc.
Power consumption, heat dissipation and noise

If you have the resources available, I'd suggest getting loaner systems from
vendors or resellers, installing your application and profiling it on each
system.

Also, be aware of the licensing situation for any third-party software
you're using.  Some products are licensed per core _installed_ versus per
core _used_.

You should also check with the vendors to see if they have any special
offerings.  IBM, for example, will ship you a system with some CPUs
disabled.  You can then activate these CPUs for certain amounts of time
should you have a spike in usage.

-Damian

ps. Clock rate is a terrible way to compare CPU throughput due to the impact
of things like word size, instruction set, pipelining, cache design, etc.
If you're interested in computer architecture, I highly recommend "Computer
Architecture: A Quantitative Approach" by Hennessy and Patterson.
=============================================================

Thanks

Levi 
 

________________________________

From: Levi Ashcol [mailto:leviashcol@hotpop.com] 
Sent: Sunday, June 18, 2006 9:21 PM
To: 'sunmanagers@sunmanagers.org'
Subject: OT: High End Servers comparison


Dear All,
  My company is in the process of purchasing a high end server for a
business critical application. My manager request a technical report
comparing the high end servers of vendors. 
 
My question:
   - What might be a "fair" points/features of comparison ? I might see
processor speed as unfair because of different CPU architecture.
   - Anybody knows an internet link/information/document for an example of
such comparison ?
 
Any clue will sure help.
 
I will summarize.
 
Thanks.
 
Levi.
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers