I was going to post that you were not allowing any memory for the o/s - then I read your post from microsoft.

I'll be blunt, you'd not configured memory correctly...

Please don't use a table variable for anything other than a small data set. You'll find a #temp table will optimise far better.

clustered indexes allow for enforced uniqueness on a...

removed properly I hope.

I really think you might need a good read of BOL.

you can set up a restricted snapshot on table columns - it's in the advanced options - don't really...

I've tried various configs for arrays and to be honest never been able to measure any difference. Tran logs and backups are best set with 64kb blocks and the data...

It always depends upon what your concept of large is .. when I've enabled deadlocking trace flags I've often ended up with 80Mb logs per day - the easisest way...

please don't tell me you're trying to populate a table variable with 15 million rows?

It sometimes makes you want to grasp them round the throat very tightly !!  I believe this issue can be resolved using a cursor which batches it's results ( forehose...

Hmmm I remember this myself - to be honest the index tuning wizard is pretty well rubbish - if you've a copy of 2005 the tuning advisor is much better. ...

I've not found that many of the perfmon counters give meaningful results with a SAN. No doubt the queue counters indicate something, but I suspect they don't relate the same...

ah! stupid idea using guids for a clustered index!! sort of fixed in 2005.

it's still in 2005.

It's just a case of trying, although I would say that SSIS has much better features susch as support for wmi so thse things should be easier.

I'll do my quick plug for sql diagnostic manager, go to http://www.idera.com and get a demo/buy a copy  and attach to your server. This will trend your box over...

You can cycle your error log   sp_cycle_errorlog  creates a new log, I normally schedule these just after midnight at a frequency which suits as required, e.g. every sunday, every night.

It should be possible but I figure it would be horrible to support, a failover in one instance would force a failover of the other I figure - could be...