tag:blogger.com,1999:blog-9144505959002328789.post1579984969571404883..comments2024-03-28T11:39:50.622+01:00Comments on Karlsson on databases and stuff: Database Innovation, pleeease!Karlssonhttp://www.blogger.com/profile/04874338187076980133noreply@blogger.comBlogger6125tag:blogger.com,1999:blog-9144505959002328789.post-6376863222957175102012-05-24T04:00:25.900+02:002012-05-24T04:00:25.900+02:00Agree with you Karlsson.
Innovation in Databases i...Agree with you Karlsson.<br />Innovation in Databases is low; the reason could be because of their dependency on storage/processors/OS and other factors.<br />I think the main areas where there is ample scope for innovation come through the processing speed, data storage etc.<br />I am in my own way trying to think of different ways of query processing, indexing etc.<br />eg: <br />Matrix databases,<br />sub columns<br />parent-child databases etcCVhttps://www.blogger.com/profile/04621878787007582589noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-88436903482480790262012-01-19T10:33:07.036+01:002012-01-19T10:33:07.036+01:00Luca!
Again, cool stuff. I'll look into t...Luca!<br /><br /> Again, cool stuff. I'll look into that. Then we need some commercial products (yes we do, to get these technologies accepted and get some money to develop them). Menawhile I'll look into these interesting ideas.<br /> But my take on the lack of innovation I think is still valid: The big database products out there are still, after all these years, awfully traditional, i.e. the likes of Oracle, MySQL, DB2 etc. It has gotten to the point that when you talk about "INDEX" in the context of a database, everyone assumes you mean a B-Tree, and most other index types are explained in terms of how they differ from a B-Tree. Like exlaining what Lada GaGa is about by talking about what is different from Manfred Mann.<br /><br />/KarlssoonKarlssonhttps://www.blogger.com/profile/04874338187076980133noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-75401496429725110822012-01-18T21:20:52.411+01:002012-01-18T21:20:52.411+01:00Innovation? Look at the MVRB-Tree of OrientDB: it&...Innovation? Look at the MVRB-Tree of OrientDB: it's derived from the well known RB-Tree (super fast in memory) but storing X entries per page like B+Tree. On memory reclaim sub trees are kept in memory.Luca Garullihttps://www.blogger.com/profile/04523569178463279505noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-21039664309199740742012-01-18T19:29:47.748+01:002012-01-18T19:29:47.748+01:00And as for LSM / Fractal trees! Fine, cool! Let me...And as for LSM / Fractal trees! Fine, cool! Let me have some real products to try them on!Karlssonhttps://www.blogger.com/profile/04874338187076980133noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-9659872488844836352012-01-18T19:28:09.297+01:002012-01-18T19:28:09.297+01:00Agreed to an extent, although for in-memory data, ...Agreed to an extent, although for in-memory data, despite the shortcomings, T-Trees are good for in-memory data. But not always. B-Trees is a technologu that has served us very well for many MANY years, but now is the time for change. And I do not imply that T-Trees is the answer, just that we need something better (and not even I think that T-trees is the answer, just that they are different).<br /><br />Thanx for commenting<br />/KarlssonKarlssonhttps://www.blogger.com/profile/04874338187076980133noreply@blogger.comtag:blogger.com,1999:blog-9144505959002328789.post-30576531839710099022012-01-18T13:06:24.431+01:002012-01-18T13:06:24.431+01:00T-trees are not as good as B+ tree even in main me...T-trees are not as good as B+ tree even in main memory. That's because even in main memory caching effects are important. One problem with T-trees is that they store data in internal nodes of the tree. The space at the root of the tree would be put to better use holding pivot keys instead of data. The wikipedia page for T-trees cites several articles which studied this problem and concluded that T-trees do not beat B-trees.<br /><br />If you want something innovative, consider LSM trees. If you want something really innovative, consider fractal trees which have the insertion performance of LSM trees without the read penalty of LSM trees.Bradley C. Kuszmaulhttps://www.blogger.com/profile/16428429628432728830noreply@blogger.com