SAP/Sybase IQ

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

SAP/Sybase IQ

JanEmilLarsen
Any thoughts or plans for including SAP/Sybase IQ as a target platform for the tool?

Does the Column-Store Architecture align well with the Anchor model?
Reply | Threaded
Open this post in threaded view
|

Re: SAP/Sybase IQ

roenbaeck
Administrator
There are currently no plans for SAP/Sybase IQ, mainly because none of us have any experience with that platform. That being said, it should be quite easy to create a port for anyone having such experience. Would you like to? :)

Column-Store works well with Anchor Modeling. Most CS implementations use some form of compression of data. In order to achieve good compression, you need to find many similarities between many rows. For this reason, when a table grows large, the compression algorithm will run slower on a wide table than on a narrow table when inserting data. So, inserts in AM on a CS are about as fast as they get.

When it comes to querying, looking at the 6NF tables in AM, for example a typical attribute, it will have three columns. One with the anchor identity, one with the actual value of the attribute, and one with a metadata reference. A query in which this attribute is involved will typically use the anchor identity and the value, but not the metadata. In this respect, two of the columns may be considered "hot" and one of them "cold". This is of course advantageous when columns are split into their own storage bins. Depending on the CS implementation, you may even be able to specify that some columns are always retreived together, such as the "hot" pair described above, effectively putting them in the same bin.

Some CS implementations are also massively parallel processing (MPP) databases, which brings along a lot of other benefits when you use Anchor. See here for more information on MPP: http://www.anchormodeling.com/?p=952