hi everybody, we have an intranet with 5 hosts connected to the internet via a router/gateway on a ppp connection. the goal is to provide every host at least 1/n of the available bandwidth (downstream is important) and even more if available. I had a look at the Advanced Routing Howto and the QoS Howto, but both are far from complete and I still have some question. As far as I understood, the SFQ (Stochastic Fair Queuing) is what I want. (it is using one queue per connection, one per hosts would exactly fit my needs, but this should be okay too) and dequeues them in "round robin", thus if a queue for a connection is empty, another queue can use its bandwidth. the problem is, that this is true only for the upstream. I would need another algorithm called RED (Random Early Detect) which drops packets if the corresponding queue is about to be full, the TCP/IP Layer will think the link is not as good as expected and slow down packet sending. my questions: do all the packets dropped by RED also pass the netfilter code before? wouldn't sfq also drop packets if the queues are full, thus slowing down the link? would it make sense to use RED and SFQ in a row, or are both implementing the same mechanisms? does RED schedule based on connections or hosts? all schedulers act on the postrouting chain, true? I would even appreciate if someone could point me to some documentation.. ;) tia stefan -- "boredom is not a burden anyone should bear" - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
