Imke Cronje (1/28/2008)

---

Am busy running DBCC ShowContig at the moment. Taking quite long.

Checked out execution plan and its shows this:

Query Cost (relative to batch) : 100%

Estimated Operator Cost: 99...

We defrag to remove frag. :w00t:

Fragmentation takes several forms. Basically when data is either not on disk sequentially or is out-of-order internally on the pages is fragmentation. ...

NOLOCK is most certainly NOT the default for SELECT statements. You must explicitly state it or be running under READ UNCOMMITTED ISOLATION to get both the benefits AND drawbacks...

See here for description of DISKPART commands: http://support.microsoft.com/kb/300415. Down at the start of the CREATE section "On all MBR disks, the size or offset parameters are rounded up...

I didn't understand it to work that way. I thought the actual files that were created were all made with even block sizes (usually 64 512 byte sectors or...

That is something that almost every non-experienced (and quite a few of the experienced) SAN configuration personnel miss. However, I am not sure I understand why it has led...

1) were all files created exactly the same size and time?

2) were all files created large enough for all data and data growth or did you let autogrowth manage the...

can you please post the output from either the index physical stats DMV or dbcc showcontig? Note that showcontig is a blocking operation.

You could also use a product like ApexSQL Audit, which can handle the grunt work for you and provide other nice features as well.

Unless these are full strings you are searching for then indexing won't get you seeks (although it may allow index scans which would probably still be faster than table scans)...

There is another very insidious and often overlooked issue related to this. You can actually reference the WRONG object if you don't use schema qualifying. If my default...

If you grab some locks on a key table in the application (Invoice, primary-key-generator, etc) and then issue a waitfor delay those locks will remain held until the waitfor is...

It could work. Then again it might not. Personally I wouldn't trust what someone told me about this if it were MY production environment we were talking about....

If the load process is serial in nature, I would take Adam's suggestion a step further and simply start the post-load job using the start-job sp call as the final...

By examining the transaction of concern (and perhaps some of those around it) it could still be possible to determine if it is a process or a person, even if...