Grant Fritchey (1/11/2013)
Are the partitions on a single disk? If so, you've just added the overhead of having to filter your data but you're not benefitting from having multiple disks to offset that cost. On a guess, I'd say that's it. Always remember, partitioning may, may, increase performance but it's primarily a data management tool so it frequently doesn't help performance at all. But that doesn't matter if you're able to manage data sets in a way you never could before.
(Disk is a San (42 disks))
Most of my wondering came from;
Using a non partitioned database, should give about the same results as using ONE single partition in a partitioned database.
(Number of diskaccesses is very similar, actions are similar etc., cpu should be similar.)
The difference in Throughput though is a factor of about 10.
So there is a flaw in my reasoning, or there is a flaw in my setup.
Any help would be welcome.
At the moment I am looking in causes for these differences, but haven't found any explanation yet. (The databases are not totally the same, but the dataset I am writing is the same). Timings do reproduce very well so my conclusion is that I do not get interference from others. (On the SAN which is not completely my own and on the instance).
Ben