SQL 2016 - It Just Runs Faster: AlwaysOn Parallel Compression / Improved Algorithms
SQL Server 2016 introduces two distinct changes in the AlwaysOn transport, compression design.
- Improved compression algorithms
- Parallel compression of log block data
Compression can be performed faster, using less resource overhead and maintains compression ratios.
SQL Server 2016 also introduces parallel compression operations. The following chart outlines the performance and resource gains as tested using a TPCC stress load.
Throughput MB/s |
Average CPU utilization (secondary) |
MB sent on wire/sec |
|
Today |
82 |
17 |
35 |
2016 |
540 |
36 |
230 |
Since the primary's compressed buffers are shared among multiple secondary transports all secondary's must support the newer algorithms or the primary uses the SQL 2014 compression algorithms.
SQL Server 2016 always performances parallel decompression regardless of the compression algorithm chosen.
'It Just Runs Faster' - SQL Server 2016 combines new compression algorithms and parallelization to significantly improve AlwaysOn log transport capabilities.
Bob Dorr - Principal SQL Server Software Engineer
Zhengguo Sun
Nithin Mahesh
Fady Sedrak
Comments
- Anonymous
May 07, 2016
What are the performance improvements in 2016 for the Standard edition please? Most of the headline improvements in the last few releases are for Enterprise only. Has a detailed feature comparison matrix been made available for 2016 editions yet please? I've seen a basic one that just skims the surface but nothing in depth. Thanks. - Anonymous
June 30, 2017
This is not working in standard and is a major bottleneck. Please update sql standard 2016 to allow parallel compression as well as parallel index creation even if limited to maxdop 4 ... this would allow us to use compression features and resolve the #1 bottleneck in SQL Standard (index creation time)