TECHNOLOGY

QoS

The HSX 6000 implements a sophisticated Quality of Service (QoS) process that consists of three major traffic management components and is designed to aggregate and carry voice, video, and data traffic over a common network while maintaining the highest system performance levels possible. Traffic management operations are implemented primarily in the HSX 6000 through two (one for the ingress and one for the egress) network processors that reside on each of the USMs.

Additionally, some QoS solutions are designed into the system to manage internal traffic as it is being processed by the HSX 6000: the T3-12/OC12-1-B ASAP PHMs perform some buffer management functions, and, when packets or cells are transferred from the ingress USMs over to the egress USMs, four traffic priority levels are supported in the internal packet switch fabric of the SFM.

Techniques

QoS techniques implement classification, traffic policing, per-flow queuing, advanced buffer management, and various scheduling algorithms as described below.

Classification
The HSX 6000 classifies incoming packets or cells using the differentiated services code point (DSCP), (IPv4 and IPv6); IEEE 802.1p, an Ethernet targeted prioritization for class of service (CoS) improvement; ATM cell loss priority (CLP) bit conformity; Frame Relay discard eligibility (DE) bit; and MPLS Experimental (Exp) bit, both E-LSP and L-LSP field assignment techniques.

Traffic Policing
Each USM can police more than 64,000 flows, simultaneously, with fine rate granularity. The flow rate can be accurately policed at ranges from 32Kbps to 2.4Gbps, with any increment. Policing algorithms supported are: dual-GCRA for ATM traffic and srTCM and trTCM algorithms (both color-blind and color-aware) for frame-based traffic such as FR, Ethernet, and IP.

Per-Flow Queuing
Each USM has up to 256K queues and can support up to 128K flows — or 64K connections (two unidirectional flows per connection). Support for 12 USMs — per HSX 6000 — brings an individual flow capacity per system of 1,536K flows (~1.5 million).

Buffer Management
Buffers retain traffic until the device can catch up during the processing of network traffic. Active queue management is the process of dropping packets, when necessary. Loss priority schemes are assigned to alleviate congestion within the buffers. In the HSX 6000, two loss priority schemes are supported: tail-drop and weighted random early detection (WRED).

Tail-drop is used for real-time traffic such as voice and video

WRED is used for non-real-time traffic such as data

Queue Scheduling
Queue scheduling schemes manage congestion, and the HSX 6000 supports various scheduling algorithms that are assigned according to whether or not the link for the incoming service type is cell or frame-based.

Frame Relay and ATM links are supported by constant bit rate (CBR), variable bit rate (VBR), nrt-VBR, and unspecified bit rate (UBR) metering. Diffserv expedite forwarding (EF), diffserv assured forwarding (AF), and best effort (BE) links are supported for MPLS E-LSPs. Multiple E-LSPs can also be supported. The HSX 6000 implements hierarchical queuing and aggregation algorithms, to provide both packet fairness and service differentiation scheduling techniques. Four priority levels (P0 through P3) are used. The HSX 6000 supports various scheduling algorithms for ATM and frame-based applications such as Frame Relay, Ethernet, and EoS (x.86 and GFP).