Demartek Reports

Demartek provides real-world, hands-on research & analysis by focusing on industry analysis and lab validation testing of server, network, storage and security technologies in its ISO 17025 accredited test lab.

, President of Demartek, includes information shown here in some of his storage presentations. His public speaking schedule is also available.

If you are looking for an industry analyst who has hands-on experience with the products and technologies he speaks and writes about, please contact us.

Related Demartek Content

Principled Technologies has acquired Demartek We are pleased to announce that Principled Technologies has acquired Demartek.

Demartek Publishes Exchange Server 2003 vs 2007 I/O Comparison Summary

Updated 4 May 2012

By , Demartek President

A newer version of this page that includes information for Exchange 2010 is available at Exchange Server 2010 I/O comparison page.

Microsoft made significant changes in moving from Exchange Server 2003 to Exchange Server 2007. The primary change is that Exchange Server 2007 is a 64-bit application, which gives it much larger memory addressability. Exchange Server 2007 can use 32GB or more of memory, while Exchange Server 2003 is limited to 4GB or less of system memory. With more memory available than was available with older 32-bit systems, Microsoft changed the caching and I/O algorithms so that Exchange Server 2007 requires fewer disk reads and writes and the new ratio of database reads to writes is approximately 1:1. Previously, this ratio had varied from approximately 3:1 to 2:1. The primary goal was to reduce the number of disk reads and writes by more effective use of memory, giving users the option to use lower-performing storage while maintaining equivalent performance.

Exchange Server 2007 writes its database in 8K pages, or multiples of 8K pages and with its new I/O coalescing can perform database I/O operations in blocks up 1MB in size to further reduce disk I/O. Previous versions of Exchange Server used 4K pages and did not have the enhanced I/O coalescing functions.

The net effect is that per user, the database disk I/O operations are significantly reduced with Exchange Server 2007 and there is the capacity to put heavier user activity onto Exchange Server 2007 than was possible with Exchange Server 2003.

We are producing deployment guides for some of the technologies described in this document.

User Profiles

Note that the User Profiles and IOPS requirements for Exchange Server 2003 and 2007 are significantly different. This is due to the reduced disk reads and writes in Exchange Server 2007. The terms “light”, “medium” and “heavy” have different meanings for the IOPS values in Exchange Server versions 2003 and 2007. In addition, Microsoft has eliminated the use of these terms for Exchange Server 2010. See our Exchange Server 2003 vs. 2007 vs. 2010 I/O Comparison Summary page for details.

The following Microsoft Technet references provide more detail for each:

Exchange Server 2003 vs 2007 Comparison Summary Table
Exchange Server 2003 Exchange Server 2007
Architecture 32-bit 64-bit
Memory Usage 4GB or less up to 32GB or more
DB Read:Write ratio 3:1 ~ 2:1 ~1:1
DB Random I/O % 100% Random 100% Random
DB Page Size 4KB 8KB
I/O coalescing max. block size - 1MB
User Profiles
Light IOPS 0.5 0.11
Average/Medium IOPS 0.75 0.18
Heavy IOPS 1.0 0.32
Large/Very Heavy IOPS 1.5 0.48
Extra Heavy IOPS - 0.64