Can one mirror a web server ?

[Pete Tenereillo]

Kevin, why do call out the browser multiple A record failover process as 
"slow"? Slow relative to what? It takes between 12 and 21 seconds with IE at 
the default settings (clearly the most common, time depends on version), and 
a minute or so with Netscape (and I think FireFox). In neither case is a 
"site unavailable" type dialog box displayed to the user. It just silently 
connects. I thought that was the intent of multiple A records??? Seems to me 
it works like a charm.

The alternative multi-site failover mechanism is BGP HRI, which (assuming it 
can be made to work reliably in the first place, i.e. without any failures) 
takes 5 minutes or so for convergence after a site failure.

The "LocalDirector", BigIP, etc., won't help if the proverbial truck hits 
the site where that equipment is (i.e. those products are not appropriate, 
or intended, for multi-site failover, if the one that the user is resolved 
to is dead, another mechanism is required).


----- Original Message ----- 
From: "Kevin Darcy" <kcd at daimlerchrysler.com>
To: <comp-protocols-dns-bind at isc.org>
Sent: Wednesday, February 02, 2005 1:16 PM
Subject: Re: Can one mirror a web server ?


> barrett bonden wrote:
>
>>Can one mirror a web server ?
>>
>>The idea is to make a  failsafe  and transparent dual computer system at 
>>two
>>sites.
>>
>>This would involve (can it be done ?) one dns setting point to two ip
>>addresses ?
>>
> You can do that. It'll result in roughly equal (at the connection level,
> unrelated to actual server responsiveness and/or load) amounts of
> traffic to each web server under normal conditions, and a potentially
> s-l-o-w browser failover if one of the webservers goes down or becomes
> unavailable for whatever reason. Some very old browsers may not be
> capable of doing address failover at all.
>
> For superior load-balancing and/or failover, you'd need to get some
> specialized hardware and/or software.
>
>
>                                       - Kevin
>
>
>
>