By Robert Quimbey, Microsoft Alliance Engineer, NetApp
This week at the Microsoft TechEd convention, NetApp will be presenting new capabilities, software and hardware including many live demonstrations utilizing hardware on the show floor as well as in our lab in Bellevue, Washington. These exciting solutions involve FlexPod, Private Cloud, Windows Server 2012, and of course the big Microsoft server apps including Exchange, Sharepoint, and SQL.
Now to zero in on my favorite subject, Exchange Server. In the past year Exchange Server 2013 shipped, though migration and sizing has only been possible for a few weeks. I am particularly excited to leak a teaser about our SnapManager 7 for Exchange – which is shipping soon – and will fully support taking space-efficient Snapshots on Exchange Server 2013.
While Microsoft has included new automatic reseeding functionality in the box, most of our customers are excited about the ability to restore in minutes, regardless of database size. With the reduction in databases from 100 to 50 per server, larger databases are even more necessary than in Exchange Server 2010. Imagine never having to reseed a database through the network, or WAN; a process than consumes valuable bandwidth, and takes quite a bit of time. No one is witnessing even close to 10Gbe wire speeds on reseeding and our snap restore is a motivated and passionate “cheetah”, compared to the reseeding “snail."
Not being restrained by database size (Microsoft supports up to 16TB) and slow reseeding enables a larger database and potentially fewer database copies than a comparable JBOD or direct attached storage design. NetApp is inherently paranoid about data performing constant CRC checking and using protection such as RAID-DP our high performance RAID-6 implementation. If you don’t have to provision enough database copies to withstand one or two database copies reseeding and still provide availability, you can reduce the total number of database copies which impacts the following:
- Raw TB of storage required
- Number of servers
- Management costs
- TCO costs (power, cooling, rack space)
Most NetApp customers follow a 2+1 or 2+2 model of database copies in the primary and secondary datacenters.
The significant reduction in both transactional IOPS (user IOPS -50%), and maintenance IOPS (-66% to 80%) in Exchange 2013 vs Exchange 2010, combined with the increase in mailbox plus archive mailbox size, pushes most customers heavily into the “capacity bound” camp. When capacity bound, many features are available that can increase disk utilization on NetApp FAS storage.
- Thin Provisioning
- Space Efficient SnapShots
- Volume-move on cDOT
- Reduce DB copies
There will always be corner cases where IOPS are abnormally high due to user profile or design requirements (very small DB size) and we have new technologies such as Flash Accel, a host based cache, to increase the achievable IOPS (+15-20%) for a given disk group or aggregate.
If you are attending Microsoft TechEd in Louisiana this year, stop by the NetApp booth (#1101) for a demonstration!