|
hi
you're thinking that just a computer can handle near 1Gbit...
If not, then you will win if the process on dropping with ipt_saru is lighter
than balancing that excedent with IPVS. I'm not using it, as saru was an
highly experimental code, and I was serving around 150000 people. And I
didn't like to be kicked ;). I played with that 2 years ago (indeed, it was
installed by a mate at job).
As DR is optimized enough (and the weight of a request is lighter than the
response), for HTTP, the numbers are like this (at least in my environment)
30 MB incoming traffic (HTTP Requests), generates aproximatelly 250MB of HTTP
Traffic.
so, balancing 1Gbits would generate around ~8GBit. I don't have yet the chance
to put that anywhere ;).
What's exactly what you need?
BR,
Francisco Gimeno
> Francisco,
> Your explanation makes perfect sense and it is in-line with the
> whitepaper published, but my issue is with this statement:
>
> "So one of the -->most important thing<-- here, is that no director has
> to put the virtual-MAC in the wire, as every director has to receive the
> packet."
>
> If every director receives every packet, how can you possible support
> traffic beyond 100% of a single director? Lets say you have 5 directors
> with 1gbit interfaces each, and you have 1.5gbit of traffic coming over
> the wire. That 1.5gbit of traffic has to be mirrored on each of the 5
> directors. Unfortunately, each of the 5 directors only has a 1gbit
> link, so how can the director process all 1.5gbit of traffic?
>
> Or, can active-active LVS not increase bandwidth?
>
|
