In essence, the issue boils down to an attempt by some DB vendors to increase database efficiency by compressing columnar data and storing each column as its own storage space. The claims that these vendors maks is that, by compressing the data and "partitioning" this data, that fewer bytes must be read from storage, increasing potential thruput.
The claim is that CPU speeds have increased significantly over the last decade, while storage I/O thruput remains basically the same. By reading compressed data from vertically partitioned storage, you can reduce I/O at the expense of CPU cycles (which we have to spare)....
IMO, this is a little like the approach IBM took with the EVI index. In an EVI, each distinct value in a column generates an index, and then this index carries a "row bitmap" so that all rows where this index value is found can be flagged with a bit in the map.
Eric
-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx]On Behalf Of Booth Martin
Sent: Tuesday, September 11, 2007 4:19 PM
To: Midrange Systems Technical Discussion
Subject: Re: iSeries blog - Is DB2 Dead?
I have to admit to ignorance here. I believe I know what a vertical
based columnar data base is, but I also believe that there is a high
likelihood I have it wrong. What are the salient differences from what
I am used to working with?
DeLong, Eric wrote:
Lol, I guess some people will not be happy until everything that works well is considered "legacy"......
http://iseries.blogs.techtarget.com/2007/09/10/is-db2-dead/
Eric
midrange.com
