08.07.2014 15:07, Alexey Kovyazin wrote:

>
> Ï am not sure that Dmitry wins here, since original question is not 100%
> correct.
> Cache is always per server process - each server process has own cache,
> in all architectures, and difference is in connection/server process model.

Not really correct for the page cache, as SuperClassic has many
per-connection caches within the same process. Absolutely correct for
the temp space cache though, which is process-based.

> So, SuperClassic is like Classic in terms of usage - the more clients
> connected the more memory will be consumed.

Correct for the page cache, but not for the temp space cache (see above).


Dmitry