It's All About the IOPS, Silly…

by Michael B. Smith, MCSE/Exchange MVP

One of the common areas that people tend to overlook when designing an Exchange Server solution is the design of the disk subsystem. And if you didn’t design your disk subsystem well, it is very likely that your Exchange Server will not perform well.

In Exchange Server 2003 and in earlier versions, the read:write ratio of the Exchange database was about 4:1. In Exchange Server 2007, it’s about 1:1 for a properly configured Exchange server. This improvement comes primarily at the expense of memory - Exchange 2007 needs much more memory to be performant than did Exchange 2003.

In Microsoft’s tests, Exchange 2007 requires 73% less IOPS than Exchange 2003. However, Exchange still needs to be able to perform lots of Input and Output Operations Per Second (IOPS). In Exchange 2007, a light email user will, on average, require 0.11 IOPS. A medium user will require 0.18 IOPS. A heavy user will require 0.32 IOPS and a very heavy user will require 0.48 IOPS. Note that this is just for the mailbox database and that database I/O is completely random. The transaction log will require about three-quarters of one-half of that per user (that is, 38% of the IOPS per user).

Consider an Exchange Server with 100 users; all falling in the very heavy user range. The database IOPS is 0.48 * 100 = 48 IOPS. The transaction log IOP is 48 * 0.38 = 19 (rounding up). That really doesn’t sound like much, does it?

However, consider this: your average 7,200 RPM SATA drive will only provide you with approximately 35 IOPS. A 10K RPM SATA drive will provide you with about 50 IOPS. This should make it immediately clear that SATA is not a high-performing solution where a single drive is concerned.

For SCSI 10K RPM, the IOPS per drive value is about 85 IOPS. For SCSI 15K RPM, the IOPS per drive value is about 110 IOPS. This goes to show you that SCSI is at least twice as fast as SATA, but that still isn’t enough for large installations. So what do you do?

Enter RAID. RAID stands for Redundant Arrays of Independent Disk. You can create large pools of disk using RAID. Generally, mirrors (RAID-1) and RAID-5 (where you spend one disk on parity information) are the most commonly seen types of RAID. However, we aren’t here to talk about RAID - except how it allows us to create more highly performing disk subsystems.

Using RAID-1, read performance is equal to (IOPS * N), where N is the number of disks in the array and IOPS is the total IOPS available from a single disk. Write performance is (IOPS * N / 2). Note that N must always be an even number (such as 2, 4, 6, 8…) with RAID-1. However, the amount of disk available is (GB * N / 2), where GB is the amount of disk space available on a single disk.

Using RAID-5, read performance is equal to (IOPS * (N - 1)). Write performance is (IOPS * ((N - 1)/ 4)). Note that N can be any number higher than 2. The amount of disk space is (GB * (N - 1)).

So, once you have calculated how many IOPS you need, you can easily determine how many disks you would need in a RAID configuration. Given RAID, even SATA disk can be used to meet high IOPS requirements. And while RAID-1 gives better IOPS more quickly, it also requires more physical spindles than RAID-5.

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.

Resources:

Exchange Server 2007 Storage Technology
http://technet.microsoft.com/en-us/library/bb738146%28EXCHG.80%29.aspx 

Mailbox Server Role Storage Requirements Calculator
http://msexchangeteam.com/archive/2007/01/15/432207.aspx 

Spell Check isn’t working?

We see frequent complaints that spell check doesn't work. While "doesn't work" can mean almost anything, it's often one of two problems.

If the message body is not spell checked, even though Outlook is set to spell check on send and F7 fails it usually means you are typing in the signature area. The easiest way to identify this problem is to right click on the text. You should see the standard context menu, with cut, copy, paste and other commands. If you see a list of signatures, you're typing in the signature area. We recommend adding two dashes (--) as the first line of all signatures so you can more easily identify where the signature begins. Both Outlook 2003 (with Word editor) and 2007 are affected by this problem.

When spell check appears to work when you send a message but the message goes out with all the errors intact, it's a bug in Outlook 2007. Check for updates at Microsoft Update as this issue was fixed some time ago.

If spell check works but the custom dictionary doesn't, it may be corrupt. You can view the custom.dic file in Notepad and delete any entries that don’t look right. Otherwise, rename the file and Outlook will create a new one.

When spell check works in Outlook 2007 but not in Outlook Express, it’s because the new spelling engine in Office 2007 applications is not used by Outlook Express. See http://www.outlook-tips.net/archives/2006/20061228.htm for solutions.

Understanding BCC

This was the week for BCC questions. Actually, just two questions, asked over and over. "How can I see who the message was BCC'd to?" and "If I Reply to All, will it go to the people who were BCC'd on the original?"

Well, it's called "Blind Carbon Copy" for a reason and there is no way you will know who was BCC'd on a message you receive unless the sender tells you because the message envelope that contains the recipients is dropped at the SMTP server. Reply to All will not send a copy to BCC'd recipients, because it doesn't know if anyone was BCC'd on the original message.

Some of the confusion may stem from a bug in an early version of Outlook and one in Outlook Express that allowed recipients to see who was BCC'd on a message. These bugs were fixed years ago. There was also a bug in Exchange (also fixed years ago) that involved messages sent from the Drafts folder using WebDAV that had addresses in the BCC field.

Other frequently asked BCC questions include:

How do it see who I BCC'd a message to?
If you send a message and use the BCC field, you can open the sent message and see the BCC field or add the BCC field to your Sent folder view.

Is there a way to use a rule to move emails that were received with a BCC to a folder?
Well, you can't set up a rule that says "if my name is in the BCC field..." but you can use the process of elimination to determine if you were BCC'd: if your address is not in the To or CC field, you were BCC'd. So you'd create a rule that did not have your name in the To or CC field. Keep in mind that this rule will apply to many mailing lists and bulk email.

Can you BCC a meeting request to someone?
Yes. Add that person as a Resource attendee or forward the meeting as an iCalendar.

Recipients of a forwarded message may be able to see recipients in the Bcc box in Exchange 2000 Server or in Exchange Server 2003
http://support.microsoft.com/?kbid=894389