Official News from Microsoft’s Information Platform
Machine Learning Blog
As we started working with early adopters on the SQL Server 2014 In-memory OLTP engine (project codenamed Hekaton) one of the first scenarios we ran into involved a session management database in SQL Server. The application was a high volume website storing ASP.net Session State in the database and running into latch contention and locking problems due to the number of concurrent users supported. As this database was a central, and mission critical, single point of scale for all web traffic it had become a performance bottleneck for the business.
As we continued to work with other customers, a number had mission critical databases storing some form of session state information. Moving the critical tables and workload suffering from latch contention and therefore scale issues into the In-memory OLTP engine has produced excellent results. Below we provide further information around this common application scenario which has seen success in being migrated to In-memory OLTP.
Typically for session state related architectures there exists a large (scale-out) number of webservers with a few tables responsible for servicing the number of users on the site. The workload pattern would be:
span style="text-decoration: underline;">Bottlenecks Experienced:
em>Latch contention:In some cases latching due to monotonical incremental keys (last-page insert) was a barrier to scale, other latching on intermediate pages also became an issue. This only allowed a certain amount of user traffic to be active at one time. The ability to scale and fully utilize the hardware resources were application bottlenecks.
Throughput/Latency: In other cases, while the application requirements were still around throughput and the number of concurrent users, the latency per business transaction also had an effect on the overall experience. Here reducing latency and increasing throughput was a primary goal.
In-memory OLTP Migration Considerations:
As I mentioned we were testing with some customers with this scenario. Here are some early results on what customers have achieved:
In conclusion, do you have a session management database which is becoming the bottleneck for your application? Unable to scale-up and fully utilize your CPU and hardware resources? If so, hopefully this helps provide you with a solution!
Interested in more patterns where we have seen success with SQL Server 2014 In-memory OLTP? Stay tuned for more blogs around patterns and a follow-up whitepaper which will include more scenarios and considerations to help you migrate to SQL Server 2014 In-memory OLTP.
SQL Server 2014 CTP1 is available for download here, or you can find the index to the complete blog series here.