Thomas has been running SQL Server 2008 live in production for over 1 year as part of the Microsoft Early Adopter Program. Below is the full interview I conducted with him about his experience with SQL Server 2008.
Tell us about yourself and your role within bwin. What does bwin do?
My name is Thomas H. Grohser, and I am Senior Database Engineer at bwin Interactive Entertainment AG. bwin is the world’s largest online gaming platform that operates in 21 languages and throughout 25 core markets (http://www.bwin.ag has all the numbers and facts). My job is it to architect and implement the Microsoft SQL Server infrastructure. This also involves capacity management and planning for the future.
Tell us a little about your environment. What types of hardware/storage do you have? What are your needs for high availability? What are your biggest technical challenges with this environment?
Since we process billions of TSQL statements every day, you can understand why we don’t use the normal commodity server. After thorough testing we decided to use the Itanium 2 IA64 platform from HP (Integrity). One key advantage of this system is its scalability. On the storage side we use a combination of local SAS disks and SAN. We operate SQL Servers running on HP Superdomes with up to 64 Itanium IA64 Cores, 768MB Cache, 512GB Memory, 400 Local disks and hundreds of spindles on the SAN. The biggest challenge we face is the lack of similar systems in operation, so getting help is not always easy. Don’t get me wrong, lots of companies use these servers to consolidate hundreds of databases onto one server or run multiple instances. The real challenge is finding a database that runs a similar type of server configuration.
You are an early adopter of SQL Server 2008. How long have you been running SQL Server 2008 in production? Has it been stable? What was the biggest concern that you had during this period? How does it compare to SQL Server 2005 in performance?
For over a year now we’ve been in production with CTP5 and it has been stable from day one. We had each server up and running within two week of the release of the CTPs and kept them running at 100% uptime until they expired or RTM became available. The RTM version has been running at 100% uptime since we went online.
With regards to 2005 vs 2008 performance: on OLTP systems you cannot expect a miracle just by upgrading. You have to adapt your application to use the new features. We expect the highest performance gain from table valued parameters since we can eliminate millions of client-server roundtrips every day.
To avoid problems during the upgrading process, we carefully evaluated and chose the application and planned out our early adopting very thoroughly. We used an application that scaled out at the database level so we had several servers running the same database doing the exact same workload, all running SQL 2005. When CTP5 was available, we replaced one of the 2005 servers with the SQL 2008 CTP5 keeping the 2005 code, and replaced a second one that used the new 2008 features. When CTP6 became available, we kept the CTP5 servers and converted two additional 2005 servers to CTP6 and so on. In the end we had almost every CPT and RC version running at the same time (CTP5 already expired) this allowed us to compare the systems very effectively. Had we run into issues, the plan was to remove the server from the scale out farm and debug the problem offline.
What features really make a difference for your situation at bwin? Is there a particular feature that you thought would be very useful, but hasn’t been as useful as you were anticipating it to be? Also, is there a new feature in SQL Server 2008 that seemed “small” initially but has really become valuable to you in production?
The best bets for us on the new feature list are: backup compression, data compression and policy management. Backup compression enables us to do a 2 terabyte backup over the network in less than 40 minutes compared to 3 hours previously.
Change data capture was not as useful as we expected mostly because the storage of the change information is in the same database as the original data. If you already have a high insert/update workload on your database, adding more workload is not efficient.
The big surprise was the automatic page repair on a mirrored database. This is a feature you only learn to appreciate if you actually need it and it has saved us a lot of time.
Did you have any difficulties in upgrading from SQL Server 2005 to SQL Server 2008? Were there any surprises?
With the RTM version, none. However I should note that we only do clean installs instead of upgrades.
Is there anything that you’d like to say about SQL Server 2008 or your experience as an early adopter?
Bottom line: SQL Server 2008 is ready for heavy duty workloads.
Since you are a speaker at the upcoming PASS Community Summit, could you tell us a little more about your session and what an attendee will get out of it?
I spent the last 3 years in a great company with fantastic co-workers, architecting, planning, building and implementing one of the fastest and most reliable SQL Server solutions. Now it’s time to share some of the lessons learned and initiate interactive discussions with the PASS Community. Attendees will learn about: the importance of building a solution that is reliable, available and scalable; what criteria to use for selecting technologies; and, most importantly, how to combine multiple features of SQL Server to achieve your desired outcome and results. The session focuses on the DBA aspect of the solution and also covers the oft forgotten or underestimated issue: hardware. I will also demonstrate an implementation of all the discussed techniques used together, requiring 7 instances of SQL Server.
Thank you for taking time to share your experiences with our community. Our readers value the experiences of others within the community. I am sure that this will be very helpful.