OK, so I couldnt get to any of SQLBits other than the free Saturday, yup, they do a whole day for free. All you need do is register in time at www.sqlbits.com and then turn up. This is the fifth happening and was held at the Celtic Manor Hotel, Newport, Wales. A very impressive hotel with plenty of staff keeping all of the debris that 300+ DBAs generate while they eat bacon/egg/sausage baps, tea/coffee, cookies, croissants, fruit cake and assorted packed lunches (I did mention this was free didnt I?).
Putting the excellent facilities to one side for a while, there were five rooms where sessions were taking place and each room had 5 sessions through the day so my primary school maths tells me that there were 30 sessions to select from. Yes, 30 – because the dinner break had sponsored sessions in each of the rooms too. The full listing is here and I had a lot of awkward decisions to make on what I wanted to see.
In the end I elected to go for a performance/tuning flavoured day so, in order of attendance, I went to
>”When a query plan geos wrong” – Simon Sabin
>”Let’s Make SQL fly – a session for developers and administrators who believe in magic” – R Meyyappan
>”Using Perfmon and Profiler” – Brent Ozar
>Quest sponsored lunch showing what to monitor and how their products Foglight and Spotlight can help – Brent Ozar
>”Vital Statistics” – Christian Bolton
>”SQL Server Optimisation stuff you won’t find on Google (yet)” – Andre Kamman
Simon’s session got things off to a great start, looking at how query plans affect the performance of queries and that sometimes the plan available isnt necessarily the one that should be used and SQL needs a hand to make a different choice. Simon carried the topic very well, demonstrating his points clearly and interacting with the audience consistently. I was handicapped in this session as the goodie bag I received didnt have a pen in it so I couldnt take any notes.
After a short break the second session demonstrated a curious effect of how many rows requested in the query affects the duration of the query. Well, obviously it does but the example had a difference of 1 row and a difference in duration from 25ms to 900ms! It turns out its all to do with the data type and the SQL engine using tempdb unnecessarily. Applying the fix Ramesh showed that there was no reason to accept the slower performance. As he stressed, find out what the problem is and then the solution is easy! Contact Ramesh at www.sqlworkshops.com.
Far and away the most enthusiastic presenter of the day Brent Ozar took us through how to use free tools that are on every server already to create baselines of current performance and then how to compare these over time to measure the effects of system changes and when something spikes what the causes may be. Its all on his blog at www.brentozar.com.
Brent jumped out of this session and straight into the sponsored lunch session telling us how Quest tools can help take the hard work out of monitoring server performance and go some way to pinpointing the actual query that is causing a problem and what to do to solve it. Foglight and Spotlight are the tools they produce to help the DBA make sure their servers are performing as good as they can.
Christian Bolton took on the topic of SQL Server table statistics in the first session after lunch and explained what part statistics play in the efficiency of SQL queries, how to maintain up to date statistics and how different versions of SQL Server have different statistics. We also got an insight on how the “cost” value in execution plans is calculated and what it means.
The final session I attended was by Andre Kamman and Henk van der Valk(Unisys) to show how SQL Server 2008 scales up when pushed. On a system with 96 cores (that’s 64 quad core CPUs) and 512GB RAM they achieved some truly staggering data transfer speed to handle ETL process. SSIS out performed bcp, BULK INSERT and all other alternatives. There were some nice tweaks that they shared with us, setting the packet size in the BIOS for example gives a huge boost to transfer speed. Sadly this session was marred by the absence of a working internet connection and the live examples were replace with video from previous events.
A lot of attendees gathered for the RedGate software sponsored drinks reception and games in the bar afterwards. There was a quick prize giving process for the various competitions that were taking place at the vendors stands through the day. The lucky winners got prizes ranging from software licenses to web cams and one person won an Xbox. The driving and football games were popular, as was the Rock Band session in one of the meeting rooms, while others simply had a drink and a chat until dinner. Sadly I had to leave to drive home but will certainly be making the effort to attend next time and hopefully for more than just one day.
I would like to thanks all of those who gave up their time and talent to make this conference happen, you are a credit to the industry.
Did I mention this was all free?