White Papers

Configuration
9 Dell EMC SC Series: Microsoft Exchange Server Best Practices | CML1037
2 Configuration
2.1 Disk layout
Microsoft provides the following storage configuration best practices for Exchange 2016/2019:
RAID is often used to both improve the performance characteristics of individual disks (by striping data across
several disks) and provide protection from individual disk failures. With the advancements in Exchange
2016/2019 high availability, RAID is not a required component for Exchange storage design. However, RAID
is still an essential component of Exchange storage design for standalone servers as well as solutions that
require storage fault tolerance.
1
Determining the storage layout before the installation of Microsoft Exchange Server is an important step since
it can have direct impact on performance when using other disk solutions.
With Exchange Server 2016, due to the reduced IOPS required, Microsoft allows placement of logs and
databases on the same volume for DAG-protected databases. The Jet database (EDB) activity resembles
sequential reading and writing to 32 KB blocks. The transaction logs see 100 percent sequential reads and
writes.
The following table shows a sample disk layout based on best practices.
Drive
Recommended configuration
Contents
C:
DAS/SAN
Windows, Exchange binaries
D:
DAS/SAN
Pagefile
E:
SAN
Database 1 / Logs 1
F:
SAN
Database 2 / Logs 2
When using Exchange Server 2016/2019 Database Availability Groups (DAGs) to create a highly resilient
database infrastructure, Microsoft Preferred Architecture guidance discusses distributing three copies plus a
lagged copy across DAG members, and utilizing the same spindle on each of four servers to host an active
copy, as well as two passive copies and a lagged copy of each of the four server's databases.
1
Microsoft
design guidance is specifically for JBOD (non-SAN) environments where larger, slower, single spindles are
used to provide storage for Exchange databases. Therefore, this Microsoft guidance does not apply to SC
Series SAN volumes. This is due to the following reasons:
• Because RAID is inherent in SC Series arrays, 3-copy DAGs are not an absolute requirement for
resilience. 2-copy DAGs are more common.
• When taking volume snapshots (Replays) with Replay Manager, an additional lagged database copy
is not needed.
1
Ross Smith IV, “The Exchange 2016 Preferred Architecture,” October 2015
https://blogs.technet.microsoft.com/exchange/2015/10/12/the-exchange-2016-preferred-architecture