Hi Venks,
GridSQL is targeted towards PostgreSQL, but you can run it against other databases.
There are a myriad of configurable settings in the gridsql.config file that are mostly documented in the Admin Guide. You can also map data types, SQL statement syntax and more. The most important thing as far as performance is concerned is row shipping and loading. For that, you want to set xdb.loader.nodewriter.use_edb_jdbc_copy to false and override xdb.loader.nodewriter.template to use the particular database's bulk loading facility. For some databases that don't take STDIN as input, you'll have to write a wrapper script.
There will be a lot of things to override, like xdb.gateway.createdb, xdb.gateway.dropdb, and various other settings depending on the RDBMS behavior, like xdb.sort.trim and xdb.sort.case.sensitive, but you can get far with the existing configuration options without having to resort to modifying the code. You will probably end up needing to change 30-40+ settings to get it to work properly.
Anyway, I encourage you to consider PostgreSQL for you solution. You can use PostgreSQL's constraint exclusion partitioning and range partition into smaller subtables, and have each of these spread out across multiple nodes in the GridSQL cluster.
Good luck,
Mason
This message was edited 1 time. Last update was at 08/12/2008 10:25:24
|