Need Optimal Schedule for Snapshot and Merge Replications

I need to schedule a snapshot replication of some tables and a merge
replication of a couple other tables. I would like to know the best
way to schedule the various replication agents.
My questions are:
- Should I arrange the schedule in this order:
Step 1. Run Snapshot Agent for snapshot replication
Step 2. Run Distribution Agents: one for each branch office
Step 3. Run Snapshot Agent for merge replication
Step 4. Run Merge Agents: one for each branch office
The reason I ask is that I am under the impression that
Distribution Agent has everything to do with snapshot
replication and has nothing to do with merge replication.
Therefore, I may want to run Distribution Agent right
after Snapshot Agent for snapshot replication -- just to
group related tasks together and out of the way.
Is my understanding correct?
What is the correct order anyway?
- I have one Distribution Agent for each branch office.
Likewise, I have one Merge Agent for each branch office.
I have two branch offices; this means I have two
Distribution Agents and two Merge Agents.
Should I start the two Distribution Agents at the same time?
Should I start the two Merge Agents at the same time?
Of course, I could have separate those two Distribution
Agents in two different time slots. But I don't want to do
this in order to avoid keeping track with two different sets
of schedules.
I have a feeling that I should be able to run those two agents
at the same time because the bottleneck is the T1 line between
the central office and each branch office. Is my understanding
correct?
Thanks in advance for any info.
Jay Chan
In general you can run all agents simultaneously if you have a small number of subscribers.
You can leave the distribution and merge agents running continoulsy and they will detect the new snapshot is available for distribution and will distribute it.
Many DBAs like to
1)run their distribution and merge agents in continuous loops
2) or schedule them to restart every 10 minutes or so and have them run continoss
This makes them more resilitent to failure.
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html