|
Archives of the TeradataForumMessage Posted: Fri, 13 Sep 2002 @ 10:03:42 GMT
Hi, Sybase IQ has been designed around the concept of bitmapped columns. If you think about how Oracle uses Bitmap indexes, and then expand that to having a single index for each column, with no actual table data existing. You can then understand how they have managed to get 48TB of data into 22TB of space. Then taking the queries below into consideration, it is very easy to see that summations are going to be quicker, as the column is directly accesable and compressed into a distinct form. Because a column is in affect a Bitmap Index, any filters coming through can directly access the related rows, without performing scans. This is the same for multiple column filters, as they can be turned directly into Bitmap Matching between indexes. These 5 queries have been tailored to show the performance of Sybase IQ at is best. What it does not explain is the increased time for data loading required to perform the column bitmap generation, and certainly does not show what happens when the data changes, which if you strive for a more Active Datawarehouse it will. Anyone who has used Oracle Bitmap Indexes will know that they perform the best when just created, and performance drops as data changes, which requires regular recreating of indexes. Which makes you wonder if you have to perform the same type of operation for Sybase IQ. Regards Matthew Winter Technical Architect
| ||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
Copyright 2016 - All Rights Reserved | ||||||||||||||||||||||||||||||||||||||||||||||||
Last Modified: 15 Jun 2023 | ||||||||||||||||||||||||||||||||||||||||||||||||