Large is not *always* best when it comes to performance (but it helps) – Mailbox Performance in Exchange and Outlook 2010

Posted by Elizabeth Ayer in email, Exchange, Exchange 2010, IT Professional, Outlook, SysAdmin on 14-05-2010

Tags: , , , ,

Elizabeth AyerExchange 2010 has solved the problem of large mailboxes, right?  Not quite.  That headline message is very misleading….. In most cases, mailbox size limits have not increased at all from Exchange 2007 to Exchange 2010, or from Outlook 2007 to Outlook 2010.

What’s ‘large’?

If I ask you, ‘Does my butt look big in this?’ you know what I mean (and how to answer).  You might also think you know what I mean by, ‘Is my Exchange mailbox large?’  but by Microsoft’s definition, you’re probably wrong.

When the Microsoft Exchange team says ‘large mailbox,’ they don’t mean one that takes up a lot of space;  they mean one with a large number of messages in a single folder.

In other words, I can have 2000 messages each with a 4MB attachment, making an 8GB mailbox, and that is not large.  But if I have a mailbox with 17,135 items in the Inbox which take up 564Mb*,  that is getting large by current MS standards.

What goes wrong with large mailboxes?

According to MS, performance starts to suffer when the number of messages exceeds a certain limit, especially in ‘critical path’ folders like Inbox or Sent Items.

Importantly, Exchange performance for a single mailbox does not depend on the size of the mailbox.  In this Exchange sizing spreadsheet from MS, it shows that limits on the number of messages per folder, but no mention of total data size.

Version Items allowed in folder
Exchange 2003 5,000
Exchange 2007 20,000
Exchange 2010 100,000

Table:  recommended limits for number of items in a mail folder

So what can you expect?  You get to 20,000 messages and everything hangs?  Not at all.   Normal  functions of reading and sending mail should behave fine.

According to the article Understanding Database and Log Performance Factors, the acknowledged performance issue is very specific:

  • IF you have an Exchange 2007 server
  • AND the end user is using OWA or Outlook in online mode
  • AND this user sorts a folder with 20,000+ items differently
  • THEN the system may be unresponsive.

Sorting hammers the Exchange server, and that, according to Microsoft, is the major scalability limitation.

What, that’s it?

Even I, a serial hoarder, don’t think that 100,000 messages in my Inbox is a happy thought.   More like a doomsday scenario.

What else in Exchange 2010 facilitates the ‘large mailbox’ ideal?  According to a Microsoft White Paper, Large Mailbox Vision (where now ‘large’ means large on disk)  the main limit on mailbox size is the expense of storage.  Therefore, they are ‘solving’ the problem of large mailboxes by database format changes that mean the Exchange server no longer needs to run on high-end storage.

By removing single instancing and introducing compression, the number of I/O operations has decreased to such an extent that you can use cheaper storage for your Exchange mailboxes (I’m oversimplifying here, but only a little).

Does that reduction in I/O operations help scalability?  In most cases, no.

If your users are running Outlook in cached mode, there is basically no relationship between the I/O load on the server and the size of mailbox possible.  Current mail accounts for most of the I/O burden.  Older data sits mostly unaccessed on the server.  The spec of the client machine has a much bigger effect on the end-user experience…..

Outlook Scalability Improvements

For anyone running Outlook in cached Exchange mode, the properties of OST files, not the Exchange server, make the biggest difference for large mailboxes.

Late in Outlook 2007,  there were major changes to the OST/PST file format to improve performance and scalability (countering major degradations introduced earlier in 2007).  Although this increases the file size on disk by about 20%, the responsiveness should be worth it.

No firm guidelines on maximum number of messages exist for Outlook, but this article on performance problems promises good performance for common actions like switching folders, even for ‘tens of thousands of items in a single folder.’

MS has also released some sizing guidelines, which apply to Outlook 2007 SP1 with the February 2009 cumulative update or higher (including Outlook 2010). These sizes are about 2.5x previous size limits  for similar performance:

  • [PST/OST files] Up to 5 gigabytes (GB): This size should provide a good user experience on most hardware.
  • Between 5 GB and 10 GB: This size is typically hardware dependent. Therefore, if you have a fast hard disk and much RAM, your experience will be better. However, slower hard drives, such as drives that are typically found on portable computers or early generation solid state drives (SSDs), experience some application pauses when the drives respond.
  • More than 10 GB: This size is where short pauses begin to occur on most hardware.
  • Very large, such as 25 GB or larger: This size increases the frequency of the short pauses, especially while you are downloading new e-mail. As described above, you can use Send/Receive groups to manually sync your mail.

Bear in mind, though, that an OST can be much larger than the reported mailbox size in Exchange; according to this MS KB, “if the mailbox size is 350 megabytes (MB), the .ost file size can be 475 MB or larger.”

Another major performance consideration is whether you keep the OST/PST files on a network share.  According to a KB describing the recent Outlook 2007 improvements, ‘This configuration is not supported by Microsoft and is known to cause performance issues generally. There are no workarounds or improvements planned….’

Summary

Effectively, Microsoft is telling us that since 2007, Exchange has had no problem with large (in disk space) mailboxes.  Apart from the relatively unusual case of users with more than 20,000 files per folder, the major issue was cost, plain and simple.

On the Outlook side, there are some limitations.  As the mailbox grows past 5Gb, degradations in cached Outlook performance creep in.

Importantly, though, this is the same for Exchange 2007 and Outlook 2007 SP2 as for 2010. There is no word of any improvements to support large mailboxes in Exchange 2010 and Outlook 2010, apart from for folders with more than 20,000 messages.

I hear all the time that ‘Exchange performance is slow because this mailbox takes up <number> Gb,’ but Microsoft says this does not happen.   I would love to hear from you if you can prove them wrong!

* my actual total at the time of writing this.  I’m not proud.

Comments (9)

[...] Large Mailbox Performance in Exchange and Outlook 2010 SysAdmin Talk [...]

We have some large mailboxes on our exchange 2003 server in both senses of the word. Before we started in place archiving, we have multiple users well over 10G in size.
And then some who just didn’t understand folders. I recently dealt with one mailbox that had over 180000 mails in the inbox alone. Yes that is 180 thousand.
Load up a new client in online mode and it was near on 10 minutes (while it nearly maxed the store) before it would show any mail.

The migration to 2010 is starting where we hope to use in user based archive folders to auto move mail and reduce the OST sizes.

180,000 in one folder in Exchange 2003 has got to be some kind of record. I’m impressed that Exchange functions at all (and also impressed that you have users with that kind of patience).

Hopefully, Exchange/Outlook 2010 will be a win for you in every way.

I believe the problem here is in the whole approach MS has taken: I just got a warning from my corporate mail server that i reached 20000 emails and that’s just from 4 months, including july and august, which are low traffic because of vacations.

About not understnading folders, well, who said that we need to use folders at all?

My email mail client is evolution, which is perfectly happy to keep everything in 1 folder and let me run searches and keep virtual folders (equals no duplication of same message in different folders).

Virtual folders allow me to radically alter the way i see my inbox without moving around a single mail.

Now i don’t mean to troll, but Exchange(s) should not only be compared amongs themselves, but also against serious unix programs.

That would give certainly a better perspective about unreasonable limitations.

I know this is an old thread, but I feel you might be interested in my recent experience in this area. I am an messaging engineer for a largish corporation, and we just had degraded performance on a CCR clustered Exchange server. Not only did the Exchange mailbox server lose performance, but all of the CAS servers (load balanced in an array) did as well – locked up connections weren’t getting released, which made F&B across the organization inaccessible.

When we were troubleshooting the issue, we found consistent event log errors that referenced a single mailbox, named BatchProcess. We checked the mailbox and found it was way oversize – it hadn’t been able to send or receive email for quite some time. However, the size limti on this mailbox was .5 GB, with blocked send/receive at .6GB. This isn’t really all that big, so we figured we’d dig a bit further into the guts of this mailbox.

We ran the following PowerShell command to see what was going on in this mailbox:
Get-Mailbox BatchProcess | Get-MailboxFolderStatistics | where { $_.ItemsInFolder -GE 1000 }
This returned a listing of all mailbox folders that had more than 1000 items in them. We found that the Inbox had more than 120,000 items, and an attempt to clear these items out, System Cleanup\Inbox, had another 200,000 items in it. When we connected to the mailbox, the client hung – we could do little with it except click and walk away for a while.

What we saw in this mailbox was a ton of read receipts, so we decided we’d use export-mailbox to dump these to never-never land. When we started clearing these out, and got the mailbox well under .5 GB, it started receiving read receipts (again) from users around the company. Once we got all the old read receipts cleared out, the mailbox was under 8 MB and had only 3000 messages in it – all in the Inbox.

Once we got this cleared out, all the CAS issues were gone – everything was running smoothly, as if this had never happened. We figured we’d better find out who owned this mailbox, and when we did, it turned out that they never opened it. They had a batch process running on another system that sent automated messages to users across the company. Once the user read the message, a read receipt went to the unmonitored mailbox. We have told them they are going to need to police this mailbox, and then they need to modify this batch process so it doesn’t require read receipts.

This made for an interesting few days troubleshooting, I assure you. (-:

[...] a really good post from Elizabeth Ayer about the impact of folders with large amount of [...]

[...] a really good post from Elizabeth Ayer about the impact of folders with large amount of [...]

An end user’s point of view: I have 916 items in the Inbox and 1788 items in the Sent Items folder, 8 items in the Trash, no other extra folders. Microsoft Office Standard 2010. When I switched to the Conversation mode the user interface ground to a halt. Deleting a message from a conversation would take a few seconds – enough delay to become annoying. I just looked in my Gmail account: there are roughly the same number of items: 2000 in the inbox and 1000 in the sent folder. Conversation mode is default in Gmail and response of the user interface is instantaneous. What is it that Google can do but Microsoft can not???

[...] info on: http://sysadmin-talk.org/2010/05/large-is-not-always-best-when-it-comes-to-performance-but-it-helps-…   ‹ Startup Folder location on Windows 7 Tagged with: exchange, limits, mailbox [...]

Write a comment