> Not to take away from what sounds like a fun and sensible idea, but done > correctly, > the suggestions that Ask made (perbal, or an LB) don't have to be viewed as > "bottlenecks" > or a single-point of failure. Load balancers can be very good things to > have. :)
I kind-of agree. Obviously I want load balancing, it's just I don't want to have to insert yet another layer to get it. If we add another layer, that layer has to be provisioned to cope with the full bandwidth of the backends, plus we need enough spare machines at that layer to provide failover. If we instead achieve the same thing using the front-end servers running varnish, we save ourselves all of the hardware that would be required for the load balancers. Meanwhile adding load balancing to varnish should have little or no measurable performance impact, and does not require any additional hardware either. Moreover, load balancing performance automatically scales as the number of varnish caches is increased. So anyway, that's a slightly more in-depth explanation for why I I think it is good practice to avoid introducing an additional load-balancing layer. _______________________________________________ varnish-dev mailing list varnish-dev@projects.linpro.no http://projects.linpro.no/mailman/listinfo/varnish-dev
