Re: redundant squid server concepts

On Tue, 23 Jun 1998, Franklin R. Jones wrote:

> => > I've got two boxes running squid (1.1.20) as peers currently
> => > primarily for fail-over redundancy. currently each server configured
> => > to handle the other's IP in the event of failure (which requires a
> => > manual intervention). load "balancing" is handled by which subset/ip
> => > address the client is using/coming from.
> => >
> => > Anyone have a more dynamic method of handling this? Ideally it
> =>
> => A number of different alternatives exists:
> =>
> => * Put a L4 switch infront of your Squid boxes (costs some money)

> Not sure this is viable, as there are 100MB paths and a
> firewall in the mix.

Today's L4 switches can do 100BaseT ... and the firewalling is a seperate
issue. Take a look at http://www.alteon.com

> => * Have each machine monitor the other and automatically take over the IP
> => when a failure is detected
>
> Yes, I could do this (was considering it) but would rather
> have something that actually scales and load balances on the
> connection side. fail over is a concern, but so is load
> balance. Currently balance is more ad-hock than determinant.

L4 seems to be the best solution. Also have look at lbnamed ... it may be
a solution you'll consider:

http://www-leland.stanford.edu/~schemers/docs/lbnamed/lbnamed.html

but bear in mind the deficiencies of DNS and Bind :-0

> => * Probably some other as well.
>
> if you or others come up with something I'm all "ears".

Use a router, Cisco Local Director, roll-your-own ... ;-)

Cheers,

John.

----------------------------------------------------
John V Cougar | Voice: 1800 065 744
Cache Manager |-----------------------------
Telstra Internet | E-Mail: cougar@telstra.net
----------------------------------------------------
Received on Wed Jun 24 1998 - 00:20:11 MDT