One of the key features of version 16 is improved compression and I’ve been doing some comparison testing between that and version 15.
The headlines were as expected with overall compression improving by a factor of 2, and up to 5 times for character data. Version 15 already compressed data up to 90%, so these improvements in version 16 are significant – both in technical terms for the product but also as one of the foundations of the TCO arguments that drive it’s adoption (don’t let anyone make you believe storage is cheap). This will make SAP Sybase IQ even more attractive for the EDW and as near line storage for HANA.
Most of the improvements come from the NBit tokenization and the new way in which sparse character data is stored and the objective of this test was to measure the delivered improvement. However the tests also showed that primary and foreign keys (in fact any HG index) are also smaller by about 25% (although it should be noted that this test doesn’t fully explore the new Tiered HG). Another welcome observation was that loading data via insert..select was around 50% faster.
Apart from improved overall compression, the test also showed that VARCHAR is now a significantly better choice than CHAR for character data of any reasonable length – certainly where it is sparse. This is an area that is a little different from version 15.