The updates fly through the system to the dsi, and then everything slows
down. It's taking about 1-2 seconds per row to apply the updates to oracle
via direct connect. However, there is virtually no cpu load for either
repserver or dco. Also, if I apply the updates via isql --> dco it's much
faster. Applying the updates via sqlplus is fast too. I thought it might
have to do with ora delaying message outputs and repserver waiting on them
for error handling, but that sounds pretty flimsy. Anyone have any ideas?
--
Jason Webster
Is this high volume? There is typically a difference between configuring for
low latency for high volume vs. low latency for low volume. By volume, I mean
the amount of data over a given time interval. As a reference (and keep in mind
that I have no "real world" experience with this), I considering anything over
5MB of new primary database tranlog to be replicated as "high volume". I have
no idea how that relates to "real world" experiences.
Let's see what some of your DSI configurations are:
batch (and if set on, dsi_cmd_batch_size)
db_packet_size
dsi_fadeout_time
dsi_xact_group_size
dsi_max_xacts_in_group
dsi_sqt_max_cache_size
If parallel-DSI is used,
dsi_num_threads
dsi_num_large_xact_threads
dsi_serialization_method
--
Thanks,
G.Carter
mailto:greg....@homemail.com
mailto:greg....@workmail.com
http://www.webaccess.net/~greg.carter