Re: Column stores

[Seth Grimes <grimes(at)altaplana(dot)com>]

Isn't Kdb pretty much about queries on what I'll call "data vectors," 
whether streamed or stored, particularly time series?  That is, you 
wouldn't have complex queries that cross a lot of fields.

The hype around column stores is generated by the entry of industry 
luminaries into the field, attracting venture money that funds high-priced 
marketing talent and initiatives, substantiated by *likely* (even if not 
yet proven) applicability of the technology to a much broader array of 
business problems, given the data explosion, than was possible in the 
past.

No, the approach is not yet established in the mainstream even if there 
are notable successes for certain types of application.  Most observers of 
the data-warehousing world would say the same about open-source and 
open-source-based DBMSes.

Clareos (Crosscut) was a more generalized column-store that failed but 
then so did Great Bridge (not that GB targeted data warehousing).  Past 
failures don't mean that Vertica or Greenplum, for example, won't go on to 
win broader-market success.

					Seth


On Fri, 1 Feb 2008, Ron Mayer wrote:

> Luke Lonergan wrote:
> > > ...column stores...
> >
> > WRT "all the rage" - you'd have to point to more than press releases 
> > about technology to make that case.  We've beaten the "all the rage" 
> > companies in every POC so far and I haven't seen or heard of 
> > significant reference installations.
>
> I thought KX System's KDB is a column-oriented database which
> had impressive reference customers (Fidelity, BofA, JPMorgan, etc)
> with their column store database since the early 90s.
>
> But yeah - I'm curious why all the hype around new column oriented 
> databases when it seems people already determined what niche 
> applications they're great for quite some time ago.