When policing and remarking excess traffic remark to the same AF class with higher drop probability

- Example: Re-mark AF 31 exceeding traffic to AF 32 (do not re-mark AF 31 to AF 21)

In an IP service provider environment, the recommended architecture includes the deployment of a DiffServ backbone with a different overprovisioning ratio for different traffic classes. Most service providers typically classify their customer traffic into three or four traffic classes (such as controlled latency, controlled load, and best effort). At the network edge, the service provider may define additional traffic classes, such as, for example, a management traffic class for all network management traffic (such as Simple Network Management Protocol [SNMP] and Telnet traffic).

In the service provider core, where all the links are very high-speed, the typical QoS mechanisms required are LLQ/CBWFQ or MDRR (on a Gigabit Switch Router - GSR) and weighted random early detection (WRED).

At the ingress to the service provider core, the customer traffic is typically policed by the service provider to ensure that the rate of customer traffic does not exceed the contractual rate. When re-marking the exceeding customer traffic, re-mark the DSCP value to the same Assured Forwarding (AF) class, but increase the drop probability. For example, re-mark AF 31 exceeding traffic to AF 32 or AF 33; do not re-mark AF 31 exceeding traffic to AF 21 or AF 11.

Design Approach to Enabling QoS

Classification: Mark the packets with a specific priority denoting a requirement for class of service from the network. Trust Boundary: Define and enforce a trust boundary at the network edge

Scheduling: Assign packets to one of multiple queues (based on classification) for expedited treatment throughout the network; use congestion avoidance for data.

Provisioning: Accurately calculate the required bandwidth for all applications plus element overhead

Scheduling: Assign packets to one of multiple queues (based on classification) for expedited treatment throughout the network; use congestion avoidance for data.

Provisioning: Accurately calculate the required bandwidth for all applications plus element overhead

A set of QoS tools is needed to manage the delay, delay variation (jitter), bandwidth, and packet-loss parameters on a network. These QoS tools include QoS classification, scheduling, and provisioning tools.

Only after traffic is identified can traffic policies be applied. The first element of a QoS policy is to identify the traffic to be treated differently. Classification must be consistent throughout the enterprise for it to be effective. QoS classification tools mark a packet with a specific priority. This marking is determined by examining the following:

■ Layer 2 parameters such as 802.1Q class of service (CoS) bits, MAC address, and Multiprotocol Label Switching (MPLS) experimental bits

■ Layer 3 parameters such as IP precedence, DSCP, and source or destination IP address

■ Layer 4 parameters such as TCP or User Datagram Protocol (UDP) ports

At the network edge, QoS markings may be accepted or rejected. This is referred to as the trust boundary. If packets are entering the network from a source that is trusted to set the packet marking properly, the packets are coming from a trusted device and can be left marked as-is. If the packets are entering the network from a device that cannot be trusted to properly mark packets, these packets must be re-marked. This is the location of the trust boundary.

The QoS scheduling tools are used to determine how a packet exits a node. Whenever packets enter a device faster than they can exit the device (as with a speed mismatch), then a point of congestion can occur. Devices have buffers that allow for scheduling of higher-priority packets to exit sooner than lower-priority packets, which is commonly called queuing. Queuing is used only when a device is experiencing congestion, and is bypassed when the congestion clears.

The QoS provisioning tools include traffic-policing and traffic-shaping tools and link efficiency mechanisms. Traffic policers and shapers are the oldest form of QoS mechanisms. These tools have the same objectives: they must identify and respond to traffic violations. Policers and shapers usually identify traffic violations in a similar manner; however, the main difference between policers and shapers is the manner in which they respond to violations, as follows:

■ A policer usually drops or remarks excess traffic.

■ A shaper delays excess traffic using buffer to hold packets and to shape the flow when the data rate is higher than expected.

Link efficiency mechanisms are typically used on slow-speed WAN links to improve their throughput and to reduce their delays. Link efficiency mechanisms include LFI, payload compression, and header compression.

Implementing QoS is a means to optimize bandwidth, but not a blanket substitute for bandwidth itself. When the network is faced with ever-increasing congestion, a certain point is reached where QoS alone will not solve bandwidth requirements. At such a point, nothing short of additional bandwidth will suffice.

Was this article helpful?

0 0

Post a comment