Grumpy Old DBA - Beware the darkside !

Come to the UK SSUG

Support your local user group!

If you missed out on attending the UK SSUG Meeting in London last night then you missed an excellent evening, OK if you can't get to London Victoria beacuse you're in Scotland fair enough, but with the close proximity to Victoria station and underground it's an easy to get to location and you get free food and drink.
Subjects covered last night were x64 and cursors, so thanks to Christian and Eric.
There should be another meeting in London in two months time, next month should be at Reading, not quite so easy to get to except for me as I only live 11 miles from the Microsoft reading Campus!

Coming soon is SQL Bits III, this will be held in Hatfield ( where? ) 

http://maps.google.co.uk/maps?utm_campaign=en&utm_source=en-ha-emea-uk-goog-gm&utm_medium=ha&utm_term=map

Note that this Hatfield is in Herts as against the Hatfields in Doncaster, Humberside or Leominster. I understand there may be a prize just for getting there as it must be within the absolute worst part of the UK motorway system - unless you know better ! I think I sense a nearness of location to a certain MVP but I could be wrong.

Here's what it says about Hatfield ( herts )
Hatfield is a modern town, situated either side of the A1(M) motorway and just a few miles north of the M25 motorway. Hmmm .. so they just built a motorway straight though it then.
Hatfield started as an old coaching town dating back to Saxon times. The town expanded following the railway's arrival in 1850 and in the 1930s, the growth was linked to the development of the aviation industry. Famous aircraft such as the Mosquito, the Comet and Trident were built here.Hmmm ... nothing much of late then.
I did search the Hatfield events page on  www.hertfordshire.com, just in case SQL Bits was there, sadly there are no events listed for Hatfield this year.

So come along to a User Group meeting, it's always good to share experiences, problems and solutions; learning from your peers is probably the most effective method I know and no matter where you think you fit in the big picture there's always something you'll know that someone else will not, all systems and experiences are different. The most important point is that the only cost is your time.

Disappointing Review.

It's not very often I'm provoked into making a criticism of a respected publication, however there's nothing like being grumpy!
I've published a number of posts about baselines, trending and performance analysis and I've been quick to remark when published examples don't really match real world, so I was surprised to read a review of compressed backup software for SQL Server and the conclusions drawn in a leading publication.

I'm a keen user of such a product, which wasn't part of the review, and although I've only used the one, but at several client sites, I suspect that the choice of the Quest or Redgate product is more a personal one, however if one is to write a review of such a product then it would make sense to provide accurate comparisions and test in an enterprise environment.

From the description of the test "server" I'd figure the tests were carried out on a laptop with an external drive. A "large", 540Gb, database was used to be fair but because the database was "so large" the tester could not actually perform a native backup or restore as there wasn't sufficient disk space, only a single TB disk was availble, so estimated times were published as the base to compare against. Hmmm - pick a number, add 5, divide by the number you first thought of, add 15 seconds to make it look good - you get my feeling on this.
A single core 1.8Ghz processor also doesn't really cut the mustard either.

Now I may be wrong but I can't really think that the average enterprise user of Lite Speed or Redgate SQL Backup will be using a low spec single processor with two sata drives - but feel free to comment if indeed your production databases do run on such a system.

So my point is that the average Production Server is likely to have multiple cores and at least half way decent scsi/fc storage, the product I use can be configured to use threads, buffers and compression ratio's - I spent some time testing which combination of these settings gave me the best compression vs time performance, and I didn't even touch the use of multiple backup files, all the types of things you'd investigate if you were looking to back up "large" databases or you want really fast backups and restores. And if you were going to write a review aimed at Database Professionals who'd be looking into deployment in the enterprise.

Such is the same with performance testing or trending, you must have a valid baseline and you must have a test environment which at least gets somewhere near to the actual production environment you intend to deploy to.

My personal view is such reviews do not add any value and also give a misguided view of the technology and products, I obviously also disagree with the recommendations of the review, but that's another matter.

Disk Partition Alignment ( SANs and Diskpart )

• I’ve recently had the experience of being involved in the migration of storage for a clustered production system from one storage unit ( the physical tin ) to another on the SAN. ( It’s a bit of a misnomer to say migrating from one san to another  <grin> )
• When I viewed the configuration document from the vendor I noted there were no references to HBA buffers, block size for the formatting or partition alignment.
• I asked a number of questions and there was some debate as it was assumed that as the vendor had been engaged to produce the upgrade and deployment document they would have configured the storage optimally for SQL Server.
• During the process it became necessary to involve Microsoft and their engineer was asked about partition alignment, he confirmed that yes it was important and the performance difference could be around 25%. 
• This confirms a couple of previous experiences before I had been on my SNIA courses and had learned about partition alignment.
• Interestingly there was a series of posts on SQLServerCentral concerning this very subject just before our migration. I’ve put the post contents below, with names removed, I’d used this as part of my evidence that our provider should align partitions on the migration. 

• I think the important point here is to always ask the question, regardless of who the Vendor might be.
• NEVER ASSUME IT’S RIGHT !