Load Balancing depending on a weight

[Barry Margolin]

In article <dfutlg$28u8$1 at sf1.isc.org>, Kurt Boyack <kboyack at gmail.com> 
wrote:

> On 9/10/05, Mark Frank <mark at mark-and-erika.com> wrote:
> > * On Sat, Sep 10, 2005 at 01:13:51AM -0400 Barry Margolin wrote:
> > > I'm pretty sure he was talking about client-side caches.  That's why he
> > > said "caching resolvers", not "caching servers".  It doesn't matter wha=
> t
> > > the caching server does if the stub resolver performs its own caching.
> 
> It shouldn't matter much if the client caches it, because it will
> already have a random address. I'll bet it would be pretty even across
> all IP addresses. You might also be able to use NAT to get another
> address pointing to the "better" web server.

I guess it depends on whether the pattern of access is lots of different 
clients or long stretches of accesses from a few clients.  The fewer 
clients there are, with each of them caching the address, the more 
likely it is that the load will be unbalanced.

But if you're not dealing with a large number of clients, I think it's 
unlikely that load balancing would even be necessary -- servers are so 
powerful these days that it's difficult for just a few users to overload 
one machine.

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***