Ok, I am making a lot of assumptions (INT instead of VARCHAR(50) being one of them) with this answer, so feel free to correct me if needed. The problem with option B is that it introduces a new join to relate Users to Alerts without any real added benefit. If joining on the UserID, it is best to index the UserID, so you can utilize seeks for your joins.
For Option A, UserID will be the clustering key (index key for the clustered index) on the Users table. UserID will be a nonclustered index key on Alerts table. This will cost 16 bytes per Alert.
For Option B, UserID will be the clustering key on the Users table. UserId will probably be the clustering key in UserMap too, to make joining more efficient. UserKey (assuming this is an INT) would then be a nonclustered index key on the Alerts table. This will cost 4 bytes per Alert. And 20 bytes per UserMap.
Looking at the big picture, one relationship, for Option A, costs 16 bytes of storage, and involves 1 join operation. Whereas, one relationship, for Option B, costs 24 bytes of storage, and involves 2 join operations.
Furthermore, there are a possibility of 340,282,366,920,938,000,000,000,000,000,000,000,000 uniqueidentifiers and only 4,294,967,296 INTs. Implementing a uniqueidentifier to INT map for a this type of relationship could cause unexpected results when you start reusing INTs.
The only reason for creating this type map table, is if you plan on creating a Many to Many relationship between Users and Alerts.
Taking all of this into consideration, I would recommend Option A.
I hope this helps,
Matt
On the face of it, this looks like a classic lookup deadlock. The essential ingredients for this deadlock pattern are:
- a
SELECT query that uses a non-covering nonclustered index with a Key Lookup
- an
INSERT query that modifies the clustered index and then the nonclustered index
The SELECT accesses the nonclustered index first, then the clustered index.
The INSERT access the clustered index first, then the nonclustered index. Accessing the same resources in a different order acquiring incompatible locks is a great way to 'achieve' a deadlock of course.
In this case, the SELECT query is:
...and the INSERT query is:
Notice the green highlighted non-clustered indexes maintenance.
We would need to see the serial version of the SELECT plan in case it is very different from the parallel version, but as Jonathan Kehayias notes in his guide to Handling Deadlocks, this particular deadlock pattern is very sensitive to timing and internal query execution implementation details. This type of deadlock often comes and goes without an obvious external reason.
Given access to the system concerned, and suitable permissions, I am certain we could eventually work out exactly why the deadlock occurs with the parallel plan but not the serial (assuming the same general shape). Potential lines of enquiry include checking for optimized nested loops and/or prefetching - both of which can internally escalate the isolation level to REPEATABLE READ for the duration of the statement. It is also possible that some feature of parallel index seek range assignment contributes to the issue. If the serial plan becomes available, I might spend some time looking into the details further, as it is potentially interesting.
The usual solution for this type of deadlocking is to make the index covering, though the number of columns in this case might make that impractical (and besides, we are not supposed to mess with such things on SharePoint, I am told). Ultimately, the recommendation for serial-only plans when using SharePoint is there for a reason (though not necessarily a good one, when it comes right down to it). If the change in cost threshold for parallelism fixes the issue for the moment, this is good. Longer term, I would probably look to separate the workloads, perhaps using Resource Governor so that SharePoint internal queries get the desired MAXDOP 1 behaviour and the other application is able to use parallelism.
The question of exchanges appearing in the deadlock trace seems a red herring to me; simply a consequence of the independent threads owning resources which technically must appear in the tree. I cannot see anything to suggest that the exchanges themselves are contributing directly to the deadlocking issue.
Best Answer
Deadlocks will always happen at some point under the default locking strategy. However, it is unlikely in your given scenario because of the straight table scan.
However, it's more likely if you have several concurrent INSERTs and SELECTs.
Using NOLOCK means dirty reads and isn't best practice. Everyone seems to suggest them though...
The alternative is to use snapshot isolation modes: the SELECT will read the last committed data, rather than being blocked by the INSERT.