White Papers

BP1013 Best Practices for Enhancing Microsoft Exchange Server 2010 Data Protection and Availability 37
when the Replication service was executing the network transfer and also reading from the source
volume.
Figure 15 DAG seeding (one database) over the replication network: source volume impact
On the target we verified, as expected, that the Replication Service executed the vast majority of disk
writes, while the Information Store had a reduced impact because the target database was
dismounted at the time of the copy until the end of the database copy process.
Figure 16 DAG seeding (one database) over the replication network: target volume impact
The last area we analyzed to fully understand the outcome was the network utilization pattern. The
Exchange Replication dedicated 1GbE network was consistently utilized at close to 50% of the
bandwidth available for the entire database copy process, and then was utilized nearly 30% during the
multiple index files copy process.
The analysis of the seeding activity over the network clearly showed a considerable impact on the
local host resources. We illustrated not only the bandwidth consumption on the replication network
but also the CPU cycles utilization on both nodes and disk access on the source copy.
7.4.2 Testing DAG seeding supported by the SAN
When we came to the scenario where the seeding process could be helped by using a SAN based
snapshot, we realized we had two different situations to address. We ran tests again with both user
load and no user load, but there was no substantial difference for the performance at the SAN level.