|
sunitabeck
Flowing Fount of Yak Knowledge
5152 Posts |
 Posted - 10/01/2012 : 08:36:37
|
I don't really know the answer to whether partitioning will help, but these are my thoughts/questions
When you said two updates clash and one of them fails, did you mean that one of the updates times out, or there is a deadlock, or something else?
Assuming it is time-out, partitioning may or may not help. Unless you are explicitly locking the tables, I would infer that it is lock escalation that results in table-level locks. If that is the case, partitioning may help. But that may also bring with it higher likelyhood of deadlocks where there may have been none before partitioning. I am saying this because it is possible that two different updates need to lock the hobts for the same set of partitions, but might go about it in different order.
So, like I said earlier, I don't know if partitioning will help or hurt. But, before you do the partitioning, it would be best to get a detailed understanding of what the updates do and why/how they are failing - whether timeout is the issue, whether the timeout interval can be increased, whether the updates can be cut up into smaller pieces so a single update wouldn't keep the tables locked etc. |
 |
|
Table Partition question!
