We have noticed that whenever we deploy a package to a new server, one where the package has never existed before, the execution times are MUCH MUCH longer.
A set of simple bulk loads (truncate + full load) take 30+min to run the first time and then take 2-3min the 2nd run onwards
And the execution times of the individual containers are lower than the time of the parent.
ie
The initialise step does a simple single row select from a tiny table and changes the last run date
Note how the individual times don't add up to the container time
Same with the main sequence container. A truncate call against a single table that no other process is touching that takes 42 sec?
The actual data flow only took 4 sec but the data flow container ran for 30sec?
Note that there's no locking and all subsequent runs of these packages take the expected amount of time.
the total run time of the above package was 5:51min, over a minute longer than the sum of the steps.
Has anybody seen this before? Where would I start looking to diagnose this?
Adelaide, Australia