From: Michael C. <cle...@gm...> - 2009-05-22 12:39:55
|
Thanks for your responses. We would certainly be undertaking our own performance testing before deciding to go this route - just hoping to ascertain some theoretical limits before doing that. In terms of the size and shape of our data, much of it is quite nested (down to maybe 7 or 8 levels) documents of moderate size (avg 10-20K), but the really high volume stuff tends to be quite flat and small. We could of course put the flat data into a relational store, leaving just the hierarchical stuff in eXist - but ideally we'd want to be able to run XPath/XQuery consistently across the whole data set. We would also not be looking to do update-in-place, but rather implement copy-on-write semantics, so that multiple versions of a document would exist in the store at any one time. The ability therefore to run queries across the latest versions of each document (or indeed at a given point in time) is important without significant degradation in performance. I did read in your article that multiversion document support was proposed but gather this has not yet been implemented - is this still on the cards? It also looks as though it is not possible to distribute the data with eXist - are there any future plans to support this? Having all the data in one place would, I imagine, impose a fundamental limit to scalability. Michael 2009/5/22 Wolfgang <wol...@ex...> > It is based on eXist 1.0, which was very different from eXist 1.2 or even >> 1.3 >> > > I forgot to say: eXist 1.0 still used the old indexing scheme (described in > my old article) which put a limit on the maximum document size to be > indexed. There has been a radical switch in the core architecture between > versions 1.0 and 1.1. The 1.1 release was the first one based on a dynamic > numbering scheme, which does not limit the size of a document (at least > theoretically). With the indexing scheme, the algorithms for most core > operations have changed as well. > > Wolfgang > |