Subject Re: [IB-Architect] Classic vs SuperServer was IB/FB Lock manager failures
Author Paul Schmidt
On 23 Sep 2002 at 13:55, Ann W. Harrison wrote:

> At 12:25 PM 9/18/2002 -0400, Paul Schmidt wrote:
> > >
> > > >maybe someone could explain to those of us who are less into the
> > > >Interbase history, how we ended up with two variations of the
> > > >same theme. Namely Classic and Super Servers.
> > >
> > > The original architecture (classic) was designed for
> > > timesharing systems and VAX clusters ...
> >
> >Are process limits still an issue,
> No, but neither clusters of machines that share disk but not memory.
> Nor is lack processing power a problem.

Clusters still exist though, I know of an organization using RedBrick on a dozen HP-
RISC machines, and there there is always the Beowulf Cluster, so clustering still
exists, I know they share disk, not sure about memory though.....

> >I guess those processes need to share
> >some memory to keep from tripping over one another, and maybe that is
> >the real issue.
> Not particularly. The problem with the classic architecture is
> not that it shares too much, but that it shares too little. Each
> connection keeps its own metadata cache and its own page cache.

Which in a clustering environment may be okay, but in a system where the
processors are all in one box, then it does need to share more.

> >I also understand that SuperServer uses threads to switch back and
> >forth between queries, rather then processes. And that it's not
> >superior to classic, just different internally (guess it was named
> >by the folks in marketing).
> Given the state of computing today, I'd say that a shared server
> is vastly superior to a multi-process access method. The multi-
> process was a good solution to a problem that no longer exists.

It depends on whether modern clusters (like Beowulf) still need the process based
technology, I don't know, but it would be cool to run FB on a 256 processor Beowulf
cluster, something that you can do with Redbrick, and maybe Oracle, but definitely
not MS-SQL!

> >the
> >real issue, is what are the problems with each, and how can those
> >problems be resolved with the least amount of work.
> We should be prepared for some work

Yes, it does, I haven't seen the source code, been too busy, perhaps I can get some
time to look at it, I gotta get over to my wifes uncles place (HS connection), and get
a copy of the source.....

Paul Schmidt, President
Tricat Technologies