Windows 7 / Getting Started

---

Discontinuation of Storage Groups

In earlier versions of Exchange Server, you could create multiple databases that used a single set of log files. These collections were called storage groups and were allowed so that databases could be kept smaller and more manageable. However, because of the limited cache memory available in 32-bit versions of Exchange Server, only four storage groups with a maximum of 20 databases could be created on a single server.

Storage groups continue to exist in Exchange Server 2007; however, having multiple databases in a single storage group is generally discouraged or prohibited in many configurations. With the ability to create up to 50 storage groups and as many databases on a single server, the best practice is to never create more than one database in each storage group, but rather create a storage group for each database.

In Exchange Server 2010, storage groups have been completely eliminated. Now each set of transaction log files only supports one mailbox or one public folder database. Not only have the storage groups been eliminated, but databases are also no longer specifically owned by any single server. All of the database objects now reside within a single database container for the entire Exchange organization. In the Exchange Server versions 2000, 2003, and 2007, a database name only had to be unique within a storage group. A database could be distinguished by the server it was created on and the storage group it was created within. Because of this many companies could have rightly used the same database names multiple times across the entire Exchange organization.

In Exchange 2010 databases are no longer tied to a specific server. Therefore, each mailbox database name must be unique across the entire organization. This means that proper attention must be given to creating a naming standard that works for your organization.

Exchange Server 2003 also introduced the Recovery Storage Group feature; this feature enables an administrator to restore database backups into a special storage group for recovery. This functionality still exists within Exchange 2010; however, with the removal of storage groups it is now called a recovery database (RDB).

Performance Improvements

The overall theme for the storage changes in Exchange Server 2010 was to move away from doing many random, small, disk IOs to a few sequential, large, disk IOs. A number of changes were done to accomplish this, such as a larger page size, improved data contiguity, and more intelligent I/O operations.

The importance of optimizing I/O can be better understood by reviewing how disk technology functions. Random disk I/O is bound by the speed at which the disk head can move around and get data. The more the disk head has to move the longer it takes to read or write data. When data is read or written contiguously the disk head does not need to move between operations and can execute I/O at a much higher rate, primarily dependent on the speed at which the disk spins.

Although improvements have been made in this regard by improving the rotational speed and the aerial density of the disk platters, random reads and writes do not perform nearly as well as sequential reads and writes. The performance improvements for sequential reads and writes have far outpaced the performance improvements for that of random reads and writes. Thus by changing the format that the data is read and written in, more performance can be extracted from the same disk hardware already in use.

Optimizing this I/O opens up possibilities for you to use storage types that would have never even been considered, such as SATA disks, which generally have good sequential I/O performance, higher storage density, and lower cost than Fibre Channel or Serial Attached SCSI storage technologies, but have poorer random I/O performance. The size and speed of the database storage are typically the largest capital cost factors in an Exchange solution. In the end, optimizing Exchange to allow for the design of solutions using lower-cost SATA disks means lower-cost solutions with better performance.