[CI] Question about sparse primary indexes

[Hideaki Kimura]

BigTable's key-map is kinda sparse primary index.
Also, I think DB2's MDC (Multidimensional clustering) table is a
sparse primary index that enables efficient lookup on any (not all)
of attributes in the index.

Both MDC and BigTable is optimized for read-intensive application.
I believe it means sparse indexing is a good way for OLAP.

Alexander Rasin wrote:
> 
> Although this is a question that came up in different context (of my DBD 
> paper), I think this is the right group of people to ask :)
> 
> Consider: CIs, in some sense, are like sparse secondary indexes which we 
> think makes them cool and unique. The question is this:
> 
> Do you know of any row-based-database that supports sparse *primary* 
> indexes?  Looking through MS SQL Server, it appears that primary index 
> has to be unique, and if it isn't unique, SQLServer will make it unique 
> by extending it.
> I don't seem to remember postgresql supporting sparse primary index.
> 
> So the question is - if sparse indexes are so cool and easy to do when 
> we're dealing with a primary index and data is sorted, how come they are 
> not used?  Or is it possible to have them in postgres/MS and I just 
> don't know how?
> 
> 
> Alex
> 
> _______________________________________________
> CI mailing list
> CI at list.cs.brown.edu
> http://list.cs.brown.edu/mailman/listinfo/ci
> 


-- 
Hideaki Kimura <hkimura at cs.brown.edu>