On 30 Oct 2000, Beotel wrote:
> First of all, I do not think you should meddle with updating all
> databases. You should make one machine a master and other as slaves,
> and you should let our replication system take care of updates.
Ok, you are probably right. The reason why I didn't suggest using it
myself is because I don't have any experience using it (as far as I know,
it's pretty new) Furthermore I'm not sure how it performs if the master is
taken down for some reason. But instead of bugging you about this I better
go and read the manual before I make a too big fool of myself ;)
Another reason why the replication system in mysql wasn't considered is
that i would like to develop a generic system, which can easily be adopted
to different databases (fx. postgres), and I would therefor prefer not to
use too many unique features of a specific system.
> All you have to do is send all updates to the master while SELECT's,
> SHOW's, EXPLAINS's and other commands that do not change data should be
> sent to the least loaded server.
Yes, that's pretty much as I imagined it.
> You can measure performance of the server in several different ways,
> one being number of mysqld threads running.
Yes, I thought about something like that, perhaps combined with the
current load of the machine running the server. I imagine gathering the
statusinformation though snmp or perhaps corba, but I haven't looked too
much into that part yet.
> Regarding client - server protocol, we do have some documentation,
> which can be sent to you.
This sounds great, I would really love to get a copy of it!
Bye, Peter Korsgaard