> Sure.
>
> http:/afs/ncsa/common/doc/web/InformationServers/Conferences/CHIwww94/Start.html
> and
>
>http:/afs/ncsa/common/doc/web/InformationServers/Conferences/CERNwww94/www94.ncsa.html
Hm. Close enough, thanks. If you advertise the full path, smart Web browsers
can avoid further loading your HTTP servers and go directly to AFS.
http://www.ncsa.uiuc.edu/afs/ncsa.uiuc.edu/common/doc/web/InformationServers/Co
nferences/CERNwww94/www94.ncsa.html
- or -
file:/afs/ncsa.uiuc.edu/common/doc/web/InformationServers/Conferences/CERNwww94
/www94.ncsa.html
I was intrigued to note that when I read your Chicago slide show via the file:
URL, all of the subsidiary links also appeared as file URLs, automatically.
Neat.
N.B. Round-robin is an easy stab at load balancing, but it works best
when you have both
1. homogeneous offered load
2. identical servers
If you don't have both, you may need some feedback into the scheduler (which
is
what the DNS server is when it shuffles records). I'm guessing you have
identical servers, but it sounds like you've been struggling to force the
offered load to be homogeneous. You might find it easier just to add a
feedback loop, which can be simple since it's a pretty centralized scheduler.
Just don't make it _too_ sensitive :-}
