Unsolved
This post is more than 5 years old
2 Intern
•
126 Posts
0
2166
December 19th, 2011 08:00
Exchange 2010 BDM Throughput Discussion
Good morning community
We all know that Exchange 2010 uses Background Database Maintance (BDM) and this IO and Throughput needs to be considered when doing a Exchange sizing. The large 256 KB BDM read IO is factored into EMC IO calculations in the form of an additional EMC recommended 20% overhead factor
For the BDM throughput according to “Microsoft Exchange 2010: Storage Best Practices and Design Guidance for EMC staorage” EMC jetstress showed that BDM account for 7.5 Mb/s per Database.
And the formula to figure pout the throughput per DB is the following
Throughput Mb/s per DB=(Total IOPS per DB) X 32KB+(BDM Throuhput per DB=7.5 MB)
I was wondering on the community’s input on this. I have been hearing a lot of different way to come up with this number. I would love to hear your side and any documents you might have to support your claim.
LouisLu
161 Posts
0
December 19th, 2011 17:00
Morning. BDM accounts for 7.5 MB/s per database based on Jetstress. So the result should be trusted. Total Bandwidth increase by ~3x with BDM from 3.5MB/s per DB without, to 11MB/s with BDM enable.
The only concern is if the 7.5 MB/s per database effected by original throughput(3.5MB/s).
Lu
pgaljan
61 Posts
1
December 20th, 2011 05:00
I tend to agree with you - BDM is a developer-contrived workload. There’s no reason (and indeed no way) for them to correlate BDM workload to the production workload, and for us to size as if they do doesn’t make much sense.
This technet article says that BDM is “approximately” 5MB/s/DB.
My hypothesis is that the BDM rate is correlated to CPU clock speed. Something along the lines of “issue a 256KB read every XXX clock ticks”. This hypothesis would explain why we see it gradually increasing over time. I’ll bet if we did a comparative analysis (in production), we’d see a direct relationship between clock speed and BDM-related bandwidth. Unfortunately, this means that sizing for BDM will always be a moving target.
duckew
2 Intern
•
126 Posts
0
December 20th, 2011 05:00
Paul,
Thank you for input it was helpfull
BryanWalsh
1 Rookie
•
11 Posts
1
December 20th, 2011 05:00
For those of you that didn't see it; Ross had a blog post last week on BDM.
http://blogs.technet.com/b/exchange/archive/2011/12/14/database-maintenance-in-exchange-2010.aspx
Near the bottom he sticks this section in...
"Database checksumming can become an IO tax burden if the storage is not designed correctly (even though it's sequential) as it performs 256K read IOs and generates roughly 5MB/s per database.
As part of our storage guidance, we recommend you configure your storage array stripe size (the size of stripes written to each disk in an array; also referred to as block size) to be 256KB or larger."
duckew
2 Intern
•
126 Posts
0
December 20th, 2011 05:00
Bryan,
Thank you the article was very good. and explained a lot, thanks for sharing