3 June 2007

This is a blog item that needs to be converted to the new Bliki format

Distributed DBMS

We've been talking about the new up and coming P2P environments with high-level applicational content for years now, but where are they? It seems to me that there's only one obstacle left in terms the components required for this transition to take place in P2P space, and that's a distributed database management system. The first step was moving up from plain file content to more object-like content using metadata overlay networks or semantic overlays, but to have a fully-fledged shift of content from web into P2P we need to be able to use our current applications, and they require the ability to query the data, usually using SQL syntax. Designing a SQL-like P2P space has some extremely difficult obstacles to overcome, but never the less solutions are under development, such as the PIER project which builds querying ability into its DHT layer.

Once the P2P space is accessible via SQL, the local peers can requests active content such as PHP or Perl from the network, and execute the code locally which execute their database queries and file IO like usual, but these queries propagate back out into the space to be processed instead of to a local database server and filesystem.

• The Architecture of PIER: an Internet-Scale Query Processor
• Prefix Hash Tables (PHT's are the new generation DHT's allowing more sophisticated querying methods)
• Skip B-Tree (Next generation Distributed Search Trees)