> Just my $0.02 but if it's in MySQL then you really don't need to expire
> each one. You can write a custom script that will do this. When you
> break it down, expire is really just finding those tokens that are
> beyond the threshold where id=x and time=y. The resultant would be
> "where time=x".
Right. Are there any scripts already out there that do this?
> But even then, you would only trim it down a manageable size per user.
> Our production database for a large number of emails (but using site
> wide) is about 40mb.
What is your bayes_expiry_max_db_size? Quite a bit larger than default I
take it.
> Even if you stuck with non-MySQL based databases (suck as Berkeley DB)
> you'd still have 160gb of aggregate data files. If you truly need
> independent DB's for each user (weather file based or MySQL) I'd
> recommend building a big MySQL cluster and managing it that way. We
> currently manage a MySQL cluster (with mirrored 300gb drives and DRBD
> replication) that houses a whopping 80mb of MySQL data.
>From what I understand, MySQL cluster design is such that the data nodes keep
all the table data in memory, which would not be feasible in a 160GB
scenario...
> I don't think this helps you much, just an opinion.
I appreciate it nonetheless!
__________________________________
Yahoo! FareChase: Search multiple travel sites in one click.
http://farechase.yahoo.com
