With data burdens for organizations increasing at dizzying rates, storage management has become more important than ever. That wasn’t been lost on Microsoft in its continuing development of Exchange Server 2010. The software maker has applied a new philosophy to how the application handles storage. It takes into account the declining price of storage and the pressure to improve performance across the storage infrastructure. It embraces using direct-attached storage instead of disk arrays, continuous replication to spare servers instead of RAID or clustering and cheap disk arrays as a substitute for tape backups.

Exchange 2010, as did Exchange 2007, has improved its handling of input/output loads for a given number of simultaneous users. One way it does that is by shelving a technique for storing copies of email messages that Microsoft has used in all previous versions of Exchange. In those versions, Exchange tries to store all copies of a message at a single location on disk. That saves disk space but reduces performance. Exchange 2010 stores copies wherever there’s free space. That may eat up more space, but Microsoft felt the performance improvement was worth it. Those kinds of improvements in Exchange 2010 opens the door for IT departments to use more economical alternatives to traditional and expensive solutions, such as substituting serial attached storage for network attached storage.

While in economy mode, an IT department may be inclined to choose SATA rather than SAS drives. That may not be a wise idea in a high performance environment. SAS drives use a technology called Tagged Command Queuing (TCQ). It copies better under heavy loads than SATA drives, which use Native Command Queuing (NCQ). Both technologies improve performance by moving the management of I/O communication with a platter to the hard drive itself. The queue holding requests for hard drive action in NCQ devices, however, tops out at 31 requests. No such limitation is set on TCQ devices. In addition, SATA drives are built to a price point so they tend not to be as reliable as SAS drives. Moreover, the price differential between the two technologies isn’t significant when “peace of mind” considerations are factored in. According to Microsoft, SAS drives cost five percent more than SATA drives, yet perform 25 percent better under peak loads.

Deployment of RAID can be avoided, too, thanks to Standby Continuous Replication. It allows each server storing mail messages to be replicated on a hot-spare server. What’s more, the technology works over long distances so the hot-spare can be located at a distant location, which will make disaster recovery planners very happy. The approach can be very resilient. Microsoft, which has implemented the approach in its own operations, has found that over the three year life span of a server, an average of one drive will fail. When a drive fails, the database on that drive fails over to a replica server. According to Microsoft, its system experiences less than an hour of downtime per year.

Exchange 2010′s new storage technologies can also provide an escape hatch from total dependency on tape for backup storage. A major challenge of IT administrators is trying to cope with shrinking backup windows. As data grows, it takes longer and longer to back it up. Couple that with an expanding demand for availability due to business hours expanding from nine to five to 24/7 and some IT honchos can find themselves in a real bind. Of course, more hardware could be thrown at the shrinking backup window problem, but the kind of iron needed to meet past levels of performance can be expensive–both in terms of the kind of tape drives required and the media needed to feed the drives.

To some extent, backup problems are partially addressed by replication of data to the hot-spares. The difficulty there is there are no historic backups. If the backup on a hot-server gets corrupted, there are no earlier backups from which to restore the data. Microsoft’s answer to that was to create another level of nearline storage manned by SATA drives configured in a relatively inexpensive RAID 5 array.

The backup regimen for the array depends on an organization’s priorities. Microsoft, for instance, does an incremental backup every 15 minutes and a full backup overnight. That allows them to keep at least a week of backups on hand should something go wrong. Since even incremental backups can degrade system performance, the company makes its backups from the hot-spares. That diverts load demand from the live databases that are online and utilizes an under worked resource–the hot-spares.