in reply to SQLite vs CDB_File vs BerkeleyDB

Interesting, thanks for the benchmark. Strong showing from CDB. I'm about to be doing some similar benchmarking on a broader range of stuff (Cache::Cache, Cache::Mmap, etc.) with a focus on multi-process use (with mod_perl), and I will definitely include SQLite after seeing this. I don't know about CDB, since it doesn't allow updates and that makes it basically useless for storing state data.

I think you hurt BDB performance a bit by using the tied interface. I never use it, since it's slower than calling the methods directly.

Replies are listed 'Best First'.
Re: Re: SQLite vs CDB_File vs BerkeleyDB
by Matts (Deacon) on Mar 19, 2002 at 19:21 UTC
    We did consider not using the tied interface, but wanted to make sure we tested what we would actually be using at work, and I couldn't see people liking the non-tied API. Maybe I'll add that, but it's fairly obvious that it'll only speed things up.

    The read-only problem of CDB is frustrating. Though the speed of regenerating the entire DB seems pretty fast, so perhaps it's not so bad for our uses (where we'd be doing rebuilds in a cron job).

[reply]
Re: Re: SQLite vs CDB_File vs BerkeleyDB
by princepawn (Parson) on Mar 19, 2002 at 18:44 UTC
    Will there be any mention of Alzabo's Sync and Store classes?
[reply]
      Probably, if time permits. There are a lot of options to cover.
[reply]

In Section Meditations