1) tempdb should have only 1 log file.

2) 64K is typically the 'generically best' option for NTFS allocation unit size when formatting the partitions. Default isn't good...

1) Version 3042 is a BAD version of SQL Server 2005 to be on. You are missing some very important updates related to the SP2 debacle(s). See here:...

1) I believe you can benefit from multiple files/groups. At least consider placing indexes in one and data in another. Also consider isolating out any very large single...

Probably a question for MS to be honest with you. I could easily see it being in R2 if they cut off that branch of development with the original...

http://support.microsoft.com/kb/920739

It is truly amazing how I/O type (random/sequential) and I/O size (1K, 2K, 8K, ... 1024K, etc) can affect net I/O throughput on a storage subsystem!

Yeah - I have heard the "a drive is a drive is a drive" mantra before too. 15K RPM SCSI (or perhaps SAS) drives are the only way to...

I don't think I have met a 9-table join yet that...

1) Definitely fix the iLO bug if the machine is HP.

2) Also investigate the Large File Copy Bug in Windows 2003. It will also lead to cache flush. ...

Do not use FASTFIRSTROW hint on a Primary Key access.

1) Be careful adding a bunch of columns to an index just to make it covering. There is maintenance cost and disk storage cost to this.

2) An...

No one has yet (that I saw) asked about the distribution of the data. Indexes are useless (unless covering) if the query optimizer estimates more than roughly high single...

...but of course for write ops you have to divide by 4 for raid 5. ...

personally I think the ease of use of microsft products - the belief that anyone...

Hire me to present the findings for you!! 😀

Seriously, as a consultant I have been the "bad guy" on a number of occassions and it doesn't bother me....

John Rowan - you use SQLIO, IOMeter or another I/O testing app to prove that the drives are not serving up sufficient IOPs or MB per sec.

Let me...