Real world Server configurations

HCL Domino Domino Forum
1

I know that I will catch some flack for this post, but I figure I will try it anyway. I’m trying to get some real world mail server configurations being used today. I’m looking for the server hardware used, partitions, etc.(AIX, Solaris, Linux, etc)

Here is what I have and trying to decide on:

  1. Total of 14,000 mail users. Want to allow for about 20% growth over next several years.

  2. Users will be utilizing mail, calendaring, dwa

  3. Servers will be connected to SAN so space will not be consideration.

  4. Will be built using Domino 8

I would like to hear what others are currently using in their environment that would be similar to this setup. This is not for applications, only for mail and we will be running Lotus Domino 8. I have freedom to architect in whatever way most sense and of course cost is always factor, but I don’t want to discard anything due to cost.

Any examples would be appreciated.

2

Subject: Real world Server configurations

Quick thought on your third point. I’ve been in multiple environments where the SAN connectivity was only evaluated for storage capacity, not disk IO. Make sure you’re evaluating the disk IO needs for your servers to avoid the SAN becoming your performance bottleneck. Check the NotesBench configurations and you’ll see lots of raid 10 arrays with plenty of disk drives used in order to deliver high IO performance.

3

Subject: Real world Server configurations

Here’s a good starting point:

http://www.notesbench.org/r7report.nsf/By$ofUsers?OpenPage

4

Subject: Real world Server configurations

Hi Mike, be careful… Trying to design a mail server infrastructure based on this very limited info is like shooting from the hip.

The approach we normally take is -in summary- as follows:

  1. analyze existing demand levels (14K users: what are they doing, how much/many, light/heavy, etc)

  2. discover optimization opportunities (why migrate & service what you don’t need?)

  3. assess & solve project risks (deployment integrity checks + projected network&server loads)

  4. develop server placement scenarios and server sizing (vendor/platform choices, wintel/linux/aix)

I can share experiences from several projects similar in size to yours, including server setups/configs that are being used. (we have more than 10TB real live statistics from our customers; over 500K end users)

Ed Brill had organized a similar topic on his blog, with interesting results: http://www.edbrill.com/ebrill/edbrill.nsf/dx/who-has-the-biggest?opendocument&comments

Hope this helps you a bit.

Wouter (dot aukema at trust-factory dot com)