Hello all,

We are attempting to finalize our disk architecture for an Exchange 2007 environment with approximately 3200 users, about 1800 of which have Blackberry devices and are moderate to heavy users of both Exchange and Blackberries.

I've done my homework on IOPS and almost all of the available documentation that I've been able to access has numbers for Exchange 2003 (including RIM's performance guides), but not much of anything for Exchange 2007, in which IOPS requirements dwindle significantly.

I've seen many posts and articles refer to the number 3.64 for additional IOPS, and many articles state that a heavy BB user adds 3-4 times the IOPS requirements to calculations.

My question is this, where is the 3.64 number generated from. How is 3.64 calculated? Is it calculated as a static 3.64 additional IOPS per Blackberry Exchange user, or is it in fact a multiplier number to be used based upon some type of baseline number of IOPS? Here's an example:

1.0 IOPS per user for Exchange 2003 heavy user profile

1.0 x 2000 users = 2000 IOPS baseline

500 mailboxes x 3.64 = 1820 IOPS additional for BB users

3820 IOPS total

This comes from the following blog:

MsExchange Blog Spot Telnet25.org: BlackBerry cause additional overhead on exchange server

I am confused as to how and where exactly the 3.64 number is derived from. Is it derived from 1.0 IOPS baseline per user multiplied by 3.64 = 3.64 additional IOPS per Blackberry user? If so, then if I change the baseline IOPS number to 0.48 baseline per user, does the 3.64 additional IOPS per Blackberry user still apply as a static number, or do I have to recalculate the additional IOPS based upon the multiplier and the baseline IOPS number? For instance:

0.48 IOPS per user for Exchange 2007 heavy user profile

0.48 x 3.64 = 1.74 additional IOPS per Blackberry user

0.48 x 2000 users = 960 IOPS baseline

500 mailboxes x 1.74 = 870 IOPS additional for BB users

Or is it still 500 mailboxes x 3.64 = 1820 IOPS additional for BB users

I really need to nail down the IOPS calculation by the end of this week if at all possible, hence this post. Much thanks in advance to anyone who may be able to lend their experience!

81 reads and not one reply? Anyone? I'm hoping at least one person watching this forum has done some type of Exchange 2007 architecture work with Blackberries involved in the mix. HELP!

Not everyone reads this forum every day - hard to believe but we do all have other jobs. Sorry can't help on the original question.

Understood, however even folks with Exchange 2003, such as yourself as listed in your sig, can help. How did you go about planning the IOPS requirements for Exchange 2003 for instance? What statistics were used and how did you generate the stats in question?

I have a high priority ticket into RIM now through my SAM to see if I can get an answer to this issue, I'll certainly post any vendors answers that we receive back here...

Hey there; welcome to the forums.

Nice environment ... I'm sure it keeps you busy.

Plan IOP requirements to have a similiar / same multiplier factor with 2007 as is stated with 2003. Personally I plan for 4-6x the IOP requirement added for a BES user. The spec says generally 4, and I always plan for capactiy / future per MS & RIM.

Its not so much a heavy BB user as much as a heavy Exchange user that is also a BB user.

If you read all of Oz's post you'll see the whitepaper he references in this where RIM gathered performance data from Exchange and saw the increase in load. RIM found a 3.64 multiplier factor based on pre/post BES implementation... part of the reason I would take that as an absolute minimum is I bet RIM had their disk arrays properly configured per MS best practice and also had the OS configured per best practice ... I'd venture to guess most everyone doesn't do this. The multiplier factor is that a multiplier ... it stays static regardless if you have a default user you assign 1 IOP to or .5 or .3 ...

Also read the comments to the post ... and use that perfmon data to guage what your current environment is doing versus what RIM / MS says BES should have.

Best thing is to look at your current IOP load and go from that. Yes Exchange 2007 does a great job with helping remove load from the disks will all the extra RAM, but as I'm sure any Exchange person will tell you ... disk configuration is still the most important thing in your Exchange hardware configuration.

Figure out your current user IOP profile, compare it to what MS says / recommends for 2007 and adjust your numbers accordingly.

Sorry for not responding sooner, I was fighting my own stupidity.

What spec are you referring to that says this? The performance guides for RIM outline disk transfers/second as the measured spec, is this the same as IOPS?

Agreed. Didn't make that clear in my original post.

So the multiplier stays static, but the basis of the multiplier does not yes?

Using my previous example:

Which of these is correct? Also, I have another question. The 0.48 IOPS/second is the "very heavy" user profile used in the Microsoft and HP calculators, however this is for a cache mode user. From a BES IOPS perspective, should we be using the online mode IOPS number for the multiplier and not the cache mode number? My reasoning here is that the BES agent and retrieval process is not a cache mode configuration, it mimics an online Exchange MAPI profile connection.

In this case, my calculations would change to:

0.48 IOPS per user for Exchange 2007 very heavy user profile (cached mode) baseline

1.37 IOPS per user for Exchange 2007 very heavy user profile (online mode)

1.37 * 3.64 = 4.99 IOPS

0.48 x 2000 users = 960 IOPS baseline

500 mailboxes x 4.99 = 2495 IOPS additional for BB users

Or is it still 500 mailboxes x 3.64 = 1820 IOPS additional for BB users

Which calculation is correct? Unfortunately RIM has zero guidance on this critical planning matter, at least that I can find.

I'd love to baseline our IOPS, however that is not possible, we are in the planning stages of a Notes 6.5.4 to Exchange 2007 migration, so I have only a POC Exchange 2007 environment in production at this point in time.

No apologies necessary, I am very thankful to have found someone that was able to contribute. Thanks again!

IOPs = Input Output / Sec ... hence the same as disk transfers / sec.

The multiplier of 3.64 is what RIM has in the doc you previously mentioned ... I always add buffer though.

I would guestimate / say yes. But only if you're comparing apples to apples with regards to your users and usage.

Using your previous example:

0.48 IOPS per user for Exchange 2007 heavy user profile

0.48 x 3.64 = 1.74 additional IOPS per Blackberry user

0.48 x 2000 users = 960 IOPS baseline

500 mailboxes x 1.74 = 870 IOPS additional for BB users

would be correct.

IIRC ... the 3.64 multiplier is based off of cached users.

You have perfect logic with this ... but again, I think the numbers are multipliers of cached users.

Right ... Check out the Performance Benchmarking guide to see values for 2003 requirements. I would honestly take this data and use it to come up with what you'd need for 2003 ... then take what MSFT says 2007 needs and multiple by the 2003 factor.

Good to know, that helps!

So we agree that it is a multiplier, and not a static additional IOPS per user number then. My architects believe it is a static number of IOPS per user, and not a multiplier. This is where there is confusion for us.

OK, now I'm going to give you a real live example of one of our regional datacenter Exchange 2007 IOPS requirements analysis. We are architecting for 800 users in one of our regional DC's. 60% BB use within the user population. 800 x .60 = 480 BB users.

So, 480 mailboxes * 3.64 = 1747.20 IOPS

We calculated baseline IOPS at 435.8 for a mix of medium (20%), heavy (60%), and very heavy (20%) user profiles. Total came out to be roughly 2183 IOPS for 800 users in a CCR Exchange cluster using 600MB mailboxes, 100% user activity, 20% maintenance space for the LUN's, etc., and so forth (basically we took all of the standard settings in the Microsoft IOPS calculator). We're looking to use an EVA 8000 for back-end disk. According to the latest HP whitepaper on Exchange 2007 and EVA 8000, each 15k RPM disk will provide approximately 110 IOPS.

2183/110 = 20 disks for the active CCR node, 35 disks for the passive CCR node. Plus, a minimum of 16 disks for the tlogs.

This adds up to 71 disks for Exchange 2007 alone. 800 users/71 disks ~ 11 users per disk. That seems high to me. Given our space requirements, we need to order 300GB 15k RPM fiber channel disks. This is about 100k of cost in disk alone.

If I take the IOPS requirements for BB out of the equation completely, my disk IOPS requirements go down by a factor of ten, down to two disks for the active CCR node and three disks for the passive CCR node. Something just doesn't seem right about BB users multiplying my physical disk requirements by a factor of ten, hence my posts here.

If I use the multiplier as an actual multiplier and NOT as a static number like I did for the numbers above, then here's how it breaks out:

435.8 baseline IOPS

0.48 IOPS per user for Exchange 2007 cached mode very heavy user profile

0.48 x 3.64 = 1.74 additional IOPS per Blackberry user

480 mailboxes x 1.74 = 835.2 IOPS additional for BB user

Total IOPS required: 435.8 + 835.2 = 1271 IOPS

Or:

1.37 IOPS per user for Exchange 2007 online very heavy user profile

1.37 x 3.64 = 4.99 additional IOPS per BB user

480 mailboxes x 4.99 = 2395.20 IOPS

Total IOPS required: 435.8 + 2395.20 = 2831 IOPS

The difference between 1271 IOPS, 2183 IOPS (what we ended up using), and 2831 IOPS from a disk purchasing perspective is rather large. It's the difference between 11 disks for the active CCR node and 26 disks for the very same active CCR node, plus the cost of the passive CCR node. The difference is financially significant, hence my questions and conundrums here on this forum.

Yep, I am VERY familiar with this guide. I've pestered RIM for performance data for Exchange 2007, they don't have anything to share just yet. Once again, thanks very much for your assistance on this matter!

It is indeed a multiplier ... I believe RIM used 1 IOP as the standard user ... hence the reason it looks like a static assignment and not as much obvious as a multiplier.

for the size of what you're doing I would seriously look into a smaller disk subsystem. Not to say that the HP ones aren't great ... just for what you're using it for I would say it is overkill in price. Use either HPs direct attached arrays with 15k SCSI or SAS drives ... what do you need a SAN for? Or perhaps use one of the more inexpensive SANs ... we all know you really need to just focus on pure disk IOPs and not any cache magic that you'll get from the bigger SANs.

Once you get away from FC disks the cost plummets ... think about it.

I made a nice spread sheet to figue out my calculations.

And put in a few links of where I found the info.


As a final note, to calculate the IOPS for a particular physical disk, you will need the specification sheet from the manufacturer. On it, hunt for the number for latency (normally between 1.0 and 2.0 ms) and the number for access (normally around 15 - 20 ms). Add the two together. Divide 1,000 by that value to get IOPS for completely random access. For example, if latency is 2 ms and access is 20 ms, then (1000 / 22) is equal to 45 IOPS.