Frank Geisler, Uwe Ricken and Torsten Strauß revive the SIG - SQL Server Internals Group after a long break. As before, we focus on the SQL Server Engine only, which has been expanded with several features in the SQL Server 2019 and thus certainly offers interesting topics. If you are interested in a certain topic, just send us an email or talk to Frank, Uwe or me directly.
Design principles of the in-memory OLTP engine – Avoid CPU and memory overhead for interpreted T-SQL
In this blog, I explain why the in-memory OLTP engine is significantly faster than the traditional on-disk engine. T-SQL is interpreted code with high CPU and memory overhead when SQL Server has to compile the command before it can be executed by the engine. In-memory OLTP reduces the number of recompilations, which may lead to much faster query execution with less impact on CPU and memory.
Microsoft introduced in-memory OLTP in SQL Server 2014, advertising that queries can be up to 100 times faster. In this blog series I will explain the design principles of the in-memory OLTP engine to explain why it has the potential to be significantly faster than the traditional on-disk engine by skipping the complexity of handling the data pages in the buffer pool and totally eliminates logical and physical reads which can result in much faster query execution and less CPU impact.
This will be my last blog about the limitations of the missing index feature. This time I will demonstrate that the missing index feature will not suggest filtered indexes. Filtered indexes reduce the index size and thus the storage allocation in the data file is reduced. Rebuilding or reorganizing these indexes also requires fewer resources. So it's worth creating filtered indexes when possible. The following example will prove that the missing index feature will not consider creating filtered indexes.
The limitations of the missing index feature – It suggests creating indexes on imprecise and not persisted columns
This will be my penultimate blog about the limitations of the missing index feature and this one will be short. Before we start, let us clarify some terms so you can better understand the table we are going to create. The data types real and float are imprecise data types for use with floating point numeric data. Floating point data is approximate; therefore, not all values in the data type range can be represented exactly. Non deterministic columns are computed columns which may return different results each time they are called.