A mechanism for flexible BW allocation in packet switching
Original Publication Date: 2005-Mar-25
Included in the Prior Art Database: 2005-Mar-25
Related PeopleOther Related People:
The switching fabric (SF) is the core of a multi packet switch (PS). Most of the data traffic in the PS (in some cases all of the data traffic) passes it via the SF. Hence, the bandwidth (BW) of the SF is actually the node BW. SFs are designed for high BW perfomance with very simple packet processing. Most of the SFs can support several classes of service, within a class the BW is shared equally between the cards. In order to enable SF supporting guaranteed BW, flexible NW share is required. The following idea is a method for the SF adapter allowing flexible BW sharing. The SF adapter is a chip, which resides on each line card that is located between the line card packet processor and the SF. The SF adapter possesses virtual output queues (VOQ). VOQs are queues located in the ingress line card, there are set of queues for each output port called queue plane. The idea is a method of organization and controlling of VOQ. Figure 1 describes a plane, there is a plane per egress port. Each plane has queue per TC (Traffic Control) (1-6 in Fig. 1), WFQ (Weighted Fair Queuing) scheduler (7 in Fig. 1), strict priority scheduler (8 in Fig. 1) and rate limiters (9-11 in Fig. 1). AF1-AF4 (Assured Forwarding) are scheduled by the WFQ scheduler.