durusmail: qp: Re: Anyone Replaced Durus with MySQL or Postgres in QP?
Anyone Replaced Durus with MySQL or Postgres in QP?
2007-09-11
2007-09-11
2007-09-20
2007-09-20
Anyone Replaced Durus with MySQL or Postgres inQP?
2007-09-11
Re: Anyone Replaced Durus with MySQL or Postgres in QP?
2007-09-12
Michael Watkins (4 parts)
2007-09-13
2007-09-13
2007-09-13
2007-09-13
2007-09-13
2007-09-13
2007-09-13
Re: Anyone Replaced Durus with MySQL or Postgres in QP?
Michael Watkins
2007-09-13
* David Binger wrote on [2007-09-13 06:10:16 -0400]:
> After the ShelfStorage is packed, you should see a huge difference.
> ShelfStorage does not load the packed index at startup or ever, so
> the 100 bytes per object rule only applies to objects commited since
> the last pack.

Ah yes of course, I'm sure I read that somewhere on the list or in
the code but somehow managed to avoid testing a packed ShelfStorage
in comparison with the sql storages. Fixing that situation, table to
come.

While messing about with SQL storages (which don't gain much by
packing - not even disk space unless a VACUUM happens to deliver
that) my brain misplaced that important ShelfStorage feature.

> Since packing can happen at any time on a live server,
> this means that the memory footprint and startup time of shelf storage
> is reduced to essentially constant time:  you can have a billion objects
> in a packed ShelfStorage database and startup time will still be
> very small.

And memory footprint as well, depending of course on the application
using the db.

So it seems that a large ShelfStorage based db might be very
practical; if machine constraints are an issue, some care might be
needed during any large data load.
reply