White Papers

Configuration
10 Dell EMC SC Series: Microsoft Exchange Server Best Practices | CML1037
• The risk of database reseeding is very low when compared to a single disk JBOD. If a disk drive in an
SC Series array fails, a SAN spare is automatically used and the server and application are unaware
of any hardware change.
• The main reason Microsoft best practices put these multiple databases on a disk is to allow for faster
reseeding of the storage space from multiple sources. Since this is not a factor for SC Series
volumes, Dell EMC does not recommend this as a best practice.
Exchange 2016/2019 supports 100 databases per server with up to 16 copies of each to other mailbox
servers in a DAG organization. A best practice is to minimize the number of databases, using as few as
recovery objectives will allow. As the number of databases increases, so does the I/O required to support the
additional data streams. This can have a negative impact on the I/O load of a system.
As environments grow larger and larger, it becomes common to run out of drive letters for the volumes. For
the purpose of scalability, it may be suitable to use volume mount points (VMP) for database and log
volumes.
The following table shows a sample disk layout based on best practices using mount points.
Drive
Recommended configuration
Contents
C:
DAS/SAN
Windows, Exchange binaries
D:
DAS (if available)
Pagefile
C:\DB
SAN
Mount points
VMP
C:\DB\Database1
Database 1 / Database 1 logs
VMP
C:\DB\Database2
Database 2 / Database 2 logs
VMP
C:\DB\Database3
Database 3 / Database 3 logs
The performance enhancements in Exchange 2019 allow a greater number of users per server with improved
latency. The benefits of increased user count per server need to balance user impact during maintenance or
outages. If a server needs to be replaced or fails it will impact additional users. This needs to be factored in
when sizing databases or server.
The failover and re-seed performance of Exchange has been improved in Exchange 2019. This is due to
index location and architecture changes, as well as reseed and failover behavior.
2.2 SC Series page size considerations
By default, the SC Series array writes data to individual 2 MB pages in a hybrid or spinning drive system.
Systems with only SSD drives will use a 512 KB page size. Exchange 2016/2019 writes database pages of 32
KB each. Write concatenation will attempt to put as many I/Os together to fill a page where possible. This will
result in I/O sizes of 256–384 KB in size.
It is very likely that during one day, many pieces of the database will be accessed and changed. That being
considered, it is also important to understand that a single block being altered constitutes a change of that 2
MB or 512 KB page where it is located on the SC Series array. Potentially, that could be the only block on that
page that was changed. Updating a single email also requires updates to multiple indexes. This can result in
larger-than-expected snapshot sizes.