[wplug] OT - request for comment on an idea - UPDATE

[Chris Romano]

On 6/3/05, Poyner, Brandon <bpoyner at ccac.edu> wrote:
> 
> > Can you give some different examples of low-TTL-drawbacks?
> 
> It is possible to shoot yourself in the foot here, but you'd have to be
> a high volume site.  I recall reading about a scenario where the TTL was
> set very low and the network was dropping packets/corrupting packets.
> DNS starts out on UDP packets and if the requester does not receive a
> response or the response has been corrupted it will send another UDP
> request.  That goes on until the requester decides to switch to the
> way-more-expensive but reliable TCP packets.  If the TTL is too low you
> can actually end up getting more TCP requests than you can handle at
> once.  This is more than a theoretical situation, but if you're that
> high of a volume these days you'd be on Akamai or something similar.
> 

I just wanted to give everyone an update.  First, thanks to everyone
for all the information.  It  is/has really helped out.

It seems that I am trying to solve two issues as if it was one.  Each
issue has different solutions.

1). Lost internet connection:
 - I think the best thing here is to get a redundent connection,
preferably from a different provider.

2). Availablity in the event of equipment failure or fire in server room:
 - We could just move our web server to a datacenter
OR
 - put a server in a datacenter to act as DNS and backup webserver.
Use an IP from the datacenter (that should never go down) as the IP
for our web server and make the server an LVS to acheive the HA. If
implemented with the solution from 1). the LVS can load balance the
request between the two connections.

I am now going to research some solutions and pricing.  I will let
everyone know the results.

I also want to thank Chris DeMarco for taking the time to talk to me.

Thanks,
Chris