perlbal use case question

[Nathan Kurz]

On Tue, Jun 10, 2008 at 12:01 PM, Michael Engelhart
<mike.engelhart at gmail.com> wrote:
> If  we were using Java or something the
> architecture would be much simpler but currently that's not an option on the
> table.

How would you approach this problem if you were doing it in Java?  I'm
not really current on what's out there on that side of things.

>   I'm not  sure where to even start at this point.

If you choose to take this route, probably just reading through the
Perlbal source would be the best start   It's short and clear, and you
can probably understand it fully in a couple hours.  But as Mike
suggests, this might not be the easiest approach if you are unfamiliar
with Perl.  While the Perlbal solution could be solid, it's not going
to be trivial.

If you get rid of the Apache layer, the approach you first described
using Perlbal internal redirects doesn't seem that bad, and you might
not have to modify Perlbal at all.  The only hard part of that is
going to be coming up with processes lightweight enough to meet your
needs and load requirements.  If you are handling more than a few user
requests per second, and your external requests might take up to a
minute, this might be hard to achieve.  What are your anticipated
loads, by the way?

Nathan Kurz
nate at verse.com