Re: Caching Architectural Considerations

Steve Green wrote:
>
> Robert Kiessling wrote:
>
> > To be more specific. Say you have three networks, a.edu, b.edu and
> > c.edu. I am assuming that networks and DNS domains are closely
> > related. a.edu and b.edu shall use proxy1.xyz and c.edu shall use
> > proxy2.xyz.
> >
> > Furthermore, define CNAME aliases
> > proxy.a.edu -> proxy1.xyz
> > proxy.b.edu -> proxy1.xyz
> > proxy.c.edu -> proxy2.xyz
>
> Thats pretty much how we do it here.
>
> It works great, and gives us the flexibility of moving the cache from
> machine
> to machine. We just change the DNS entry, wait for the change to
> proliferate and then switch off the old cache.

If you want real redundancy it would be best to lower the TTL (time to live)
for the records in the DNS server. This way, if you make an update, it
won't take 8 hours for everything else to know about it. You can probably
set it to something like 600 seconds. This way the worst thing that could
happen is that some people are down for 10 minutes...

Oskar
Received on Wed Oct 30 1996 - 23:01:41 MST