The idea of writing a new in-memory engine is mentioned on ideas page
so I think there must be a reason for it to be here.
I also believe I've seen an FR for such feature on bugs.mysql.com, but I
was unable to find it now.
Anyway, I do think the RAM amount limitation to be a serious problem.
Even a small company using MySql Server could have
a relatively large database (i.e. 8-16 Gb or more) while computer
supporting such amounts of RAM isn't a cheap thing (and this is
exactly what I have now :)).
As for me, I use RAM tables for temporary processes with high
performance needs. Current in-memory engine forces me
to use fixed CHAR(xxx) fields for my tables which is no good, because it
sometimes causes those temporary tables to grow up to 1 Gb or more...
> Thanks for your interest in SoC.
> Your intention is commendable, but I have a question.
> Why writing a new in-memory engine, when you can get the same result
> now in a few minutes with MyISAM , by setting the data directory on a
> RAM disk?
> If you do that, you have a fully functional engine that can work with
> BLOB and TEXT columns. Your only limitation is the amount of RAM in
> your box.
> As for persistence, you can achieve that by replicating the database
> to a more durable storage.
> Another possibility is to use Brian Aker's memcached engine (check his
> blog for more info).
> Can you redefine your project, taking into account the above remarks?