Ok Super DBA's, I need your help. I'm not a dba, I'm a db architect and
do not know replication intricacies and need your help discerning the
difference between these two replication models and why you'd pick one
over the other.

The situation: a production db server (sql 2005 Enterprise x64) with
one active db on it, 6 active tables getting hit hard - our stress
testing went up to 6000 write operations/sec, hit maybe 30% cpu. On
startup, SQL Server took 8gb of the 16gb of ram, never needed more. It
is connected to a SAN via dual 2gb HBA cards. There were no page faults
or disk queues.

Now, after our self-taught dba applied transactional replication, with
only 60 operations/second, we're running 80-95% cpu, and most bizarre,
only 1 gb of ram is in play. There are still no page faults or disk
queues. Since this is a phased rollout, we haven't yet turned on the
big guns that will pound this server. Using the default memory
settings for sql server (no limits). Just changed them to 8gb minimum
and 14.5 gb max, no change.

So I'm wondering - would log shipping be a better route? We don't need
real time replication, just near real time - every 5 minutes would be
fine. Any other ideas about what to look for?

Mike Sofen


MS Sql Server LazyDBA home page