Marking and Re-marking

The OpenArchitect switch can mark or remark packets using the TOS field or 802.1p tag. This is also controlled through the Linux iptables utility.

Scheduling

The servicing of configured queues by the switching fabric is referred to as scheduling. The OpenArchitect switch has three built-in scheduling algorithms. The type of scheduling algorithm used is implied, rather than being explicitly specified, based on the number of queues and which options are configured. The following scheduling algorithms are provided:

First In First Out (FIFO) – When only one queue is configured per port, packets are serviced in the order in which they arrive. This is the default for the OpenArchitect switch.

Strict Priority – This algorithm is used when more than one queue is provisioned on the port. The highest priority queue, which is also the highest numbered one, is always serviced first (Example: If four queues are configured, queue three is of higher priority than queue zero). As long as there are packets in the highest priority queue, the lower priority queues are not serviced. The danger is that higher priority traffic could block lower priority traffic.

Weighted Round Robin (WRR) – This algorithm is similar to Strict Priority scheduling, but it provides fairness with quanta for each queue. Each queue is assigned a number of packets, known as weight, that it is allowed to transmit before it yields to a lower priority queue. Note that with WRR, the priorities of the queues are dependent on the weights allocated. A higher priority queue with a smaller weight will get less wire-time than a lower priority queue configured with a larger weight. The relative weights used for priority queues on a port can be set using the zcos command (this is a switch-wide parameter).

ztmd Explained

ztmd is a traffic management daemon which accepts messages from traffic filtering and quality of service applications and sets up the hardware.

zfilterd Explained

zfilterd is a daemon that intercepts filtering rules entered by the user via iptables, checks them for validity and then passes them on to ztmd for entry in the switch.

Running zfilterd

Before starting zfilterd, ztmd must be running. Your can start both from within a script, or directly from the command line. For example,

ztmd zfilterd

iptables rules can be entered at any time. If your iptables filtering rules set is extensive,

Ethernet Switch Blade User's Guide

release 3.2.2j

page 58

Page 58
Image 58
Znyx Networks bh5700 manual Marking and Re-marking, Scheduling, Running zfilterd, Ztmd Explained