As promised in part 2 of this series, I'd said I would post an update once the Domino server statistic IO.Number.Writes count was comparable between the old server configuration (with Symantec Mail Security for Domino) and the new server configuration (without SMSDOM, and using Lotus Protector as a gateway, off Domino spam solution). Well now the statistics are similar so we can now see the over all effect of moving transaction logs to a discrete drive whilst also keeping the statistics "somewhat" honest.

First a recap, here is the transaction logging stats before the move:

Image:A very un-scientific look at transaction logging on Domino - part 3 - a real world server update

As you can see above, the IO.Number.Writes is just over 433,000.


Image:A very un-scientific look at transaction logging on Domino - part 3 - a real world server update

The above took 33 days to get over 430,000 IO.Number.Writes on the new config (as opposed to 6 days with old). Still the overall reduction in both IO.Avg.Write.Time and IO.Max.Write.Time are staggering.

Conclusion, the jury is in. Move the logs. Period. See comment 10 on part 2 for details on how to move them.

Now if some hardware vendor wishes to give me a SAN I can run these tests against that configuration too. :)

Off on a tangent, this also outlines the potential performance issues with DAOS enabling the server mail.box while also using on-Domino spam engines. Specifically ones integrated into the mail.box, and using said server as a primary mail server (using a hub Domino server for spam would be far better in this scenario)The number of IO.Number.Writes increased over 5 fold on this server when it was integrated (33 days/6days=the factor increase). This maybe another reason to look at Lotus Protector. Watch for more on that in the future.

And, yes, the SMSDOM related databases were not TX logged. But the mail.box was. Here endeth the lesson.
Darren Duke   |   April 7 2010 03:05:22 AM   |    domino  performance  transaction logging    |  
  |   Next Document   |   Previous Document

Discussion for this entry is now closed.

Comments (0)

No Comments Found