Re: Webcluster session storage, was vacuum, performance, and MVCC

On Fri, Jun 23, 2006 at 10:31:22AM -0400, James Robinson wrote:
> Regarding the best place for session data, memcached isn't really the
> answer, for opposite reasons as to why it isn't so great to store it
> in the central DB for a bug web farm.

The thought that occurred to me while reading this is that you don't
what a database at all, you just want a quick tuplestore. In that case,
why not just create a wrapper around something like Berkley DB.

select * from bdb_get_key('session1234') as mysessiontype;
... session data ...

select bdb_put_key('session1234', ROW(... session data...))

select bdb_del_key('session1234');

select bdb_dump();
select bdb_destroy();

No pesky transactions, no vacuuming, (supposedly) high speed access.
Recent versions are supposed to be able to work with shared disk
storage between servers. AFAIK BDB is BSD licenced, so you could even
make a custom version that used the Postgres buffer cache and file
management.

Another possibility is something like tdb.

Basically, I'm not sure what all of this has to do with the core goal
of Postgres, which is to be an SQL compliant database.

Have a nice day,
--
Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.