Cascading

[Arslan Ghani]

HI !

Would u plz help me telling that how could we do cascading, that is to use
two proxy servers at a time, to do load balancing and provide redundancy.....

What parameters have to b changed and what else have to b done to accomplish
this task...

Regards ..

Arslan

____________________________________________________________________
Get free email and a permanent address at http://www.amexmail.com/?A=1

[Kendall Lister]

On 3 Jan 2000, Arslan Ghani wrote:

> Would u plz help me telling that how could we do cascading, that is to
> use two proxy servers at a time, to do load balancing and provide
> redundancy.....

You have a lot of flexibility in how you implement this, and as a result
there isn't really a simple list of steps to follow. You need to set up
two (or more) Squid servers, preferably each on its own machine for
simplicity and redundancy. You then need to have them sharing objects, so
a sibling relationship would make sense. Finally you need to handle
directing your clients' requests to each machine in a balanced fashion -
this could be done via some kind of DNS round-robin system, or you could
use a specific load-balancing router (either software under Linux or a
separate hardware box). Or you could use one server as a front end to
receive all the requests, and then have it use each of the redundant
servers as parents. We used this to solve the problem of having two
external net links - we used a second Squid box in "parallel" with our
primary server; it didn't cache anything, but it allowed us to utilise two
Internet pipes without any messy network level load-balancing on our
2.0.x kernel servers.

It strikes me, however, that if you are stuck at the level of "we need
load balancing - how?", then you're quite likely to have a difficult time
implementing it without a turn-key solution, which I don't believe exists
for Squid.

--
Kendall Lister, Systems Operator for Charon I.S. - ken...@charon.net.au
Charon Information Services - Friendly, Cheap Melbourne ISP: 9589 7781

[Blue Lang]

On Tue, 4 Jan 2000, Kendall Lister wrote:

>
> It strikes me, however, that if you are stuck at the level of "we need
> load balancing - how?", then you're quite likely to have a difficult time
> implementing it without a turn-key solution, which I don't believe exists
> for Squid.
>

for $1,000/day plus travel, i'll happily be a turn-key solution, as i bet
a lot of other people here would. :)

seriously, does anyone know of any companies providing that sort of
service utilizing the mighty squid?

--
Blue Lang, Unix Systems Administrator, QSP Inc. - Raleigh, NC
Phone: 919.875.6994
zsh: command not found: fortune

[mba...@journey.net]

He could always go to the devel versions and uce wccp with a cisco router.
Has a couple nice features. Will split the load though by IP not by usage.
Also if both fail it will go direct to the internet. Also is transparent
to the user..

--Matt

------------------------------------------------------------------------
Matthew S. Bailey, President Journey Communications, Inc.
mba...@journey.net PO Box 2003
(517) 779-2400 Mt. Pleasant, MI 48858

Inexpensive Wholesale Services for Michigan -- michix.com
------------------------------------------------------------------------

On Tue, 4 Jan 2000, Kendall Lister wrote:

> On 3 Jan 2000, Arslan Ghani wrote:
>
> > Would u plz help me telling that how could we do cascading, that is to
> > use two proxy servers at a time, to do load balancing and provide
> > redundancy.....
>
> You have a lot of flexibility in how you implement this, and as a result
> there isn't really a simple list of steps to follow. You need to set up
> two (or more) Squid servers, preferably each on its own machine for
> simplicity and redundancy. You then need to have them sharing objects, so
> a sibling relationship would make sense. Finally you need to handle
> directing your clients' requests to each machine in a balanced fashion -
> this could be done via some kind of DNS round-robin system, or you could
> use a specific load-balancing router (either software under Linux or a
> separate hardware box). Or you could use one server as a front end to
> receive all the requests, and then have it use each of the redundant
> servers as parents. We used this to solve the problem of having two
> external net links - we used a second Squid box in "parallel" with our
> primary server; it didn't cache anything, but it allowed us to utilise two
> Internet pipes without any messy network level load-balancing on our
> 2.0.x kernel servers.
>

> It strikes me, however, that if you are stuck at the level of "we need
> load balancing - how?", then you're quite likely to have a difficult time
> implementing it without a turn-key solution, which I don't believe exists
> for Squid.
>