Differentiated services

From Wikipedia, the free encyclopedia

Jump to: navigation, search

This article is on differentiated services within communication networks. For a related concept used as a design pattern for business applications (including smart services and context-aware services) see Differentiated service (design pattern).

Differentiated Services or DiffServ is a computer networking architecture that specifies a simple, scalable and coarse-grained mechanism for classifying, managing network traffic and providing Quality of Service (QoS) guarantees on modern IP networks. DiffServ can, for example, be used to provide low-latency, guaranteed service (GS) to critical network traffic such as voice or video while providing simple best-effort traffic guarantees to non-critical services such as web traffic or file transfers.


[edit] Background

Since modern data networks carry many different types of services, including voice, video, streaming music, web pages and email, many of the proposed QoS mechanisms that allowed these services to co-exist were both complex and failed to scale to meet the demands of the public Internet. In 1998, the IETF published RFC 2475 (An Architecture for Differentiated Services). Today, DiffServ has largely supplanted other Layer 3 QoS mechanisms (such as IntServ) as the primary protocol routers use to provide different levels of service.

[edit] Traffic Management Mechanisms

DiffServ is a coarse-grained, class-based mechanism for traffic management. In contrast, IntServ is a fine-grained, flow-based mechanism.

DiffServ operates on the principle of traffic classification, where each data packet is placed into a limited number of traffic classes, rather than differentiating network traffic based on the requirements of an individual flow. Each router on the network is configured to differentiate traffic based on its class. Each traffic class can be managed differently, ensuring preferential treatment for higher-priority traffic on the network.

The DiffServ model does not incorporate premade judgments of what types of traffic should be given priority treatment; that is left up to the network operator. DiffServ simply provides a framework to allow classification and differentiated treatment. DiffServ does recommend a standardized set of traffic classes (discussed below) to make interoperability between different networks and different vendors' equipment simpler.

DiffServ relies on a mechanism to classify and mark packets as belonging to a specific class. DiffServ-aware routers implement Per-Hop Behaviors (PHBs), which define the packet forwarding properties associated with a class of traffic. Different PHBs may be defined to offer, for example, low-loss, low-latency forwarding properties or best-effort forwarding properties. All the traffic flowing through a router that belongs to the same class is referred to as a Behavior Aggregate (BA).

[edit] DiffServ Domain

A group of routers that implement common, administratively defined DiffServ policies are referred to as a DiffServ Domain.

[edit] Classification and Marking

Network traffic entering a DiffServ domain is subjected to classification and conditioning. Traffic may be classified by many different parameters, such as source address, destination address or traffic type and assigned to a specific traffic class. Traffic classifiers may honor any DiffServ markings in received packets or may elect to ignore or override those markings. Because network operators want tight control over volumes and type of traffic in a given class, it is very rare that the network honors markings at the ingress to the DiffServ domain. Traffic in each class may be further conditioned by subjecting the traffic to rate limiters, traffic policers or shapers.

[edit] Per-Hop Behavior

The Per-Hop Behavior (PHB) is indicated by encoding a 6-bit value—called the Differentiated Services Code Point (DSCP)—into the 8-bit Differentiated Services (DS) field of the IP packet header. The DS field is the same as the TOS field, and ECN occupies the upper 2 bits.

In theory, a network could have up to 64 (i.e. 26) different traffic classes using different markings in the DSCP. The DiffServ RFCs recommend, but do not require, certain encodings. This gives a network operator great flexibility in defining traffic classes. In practice, however, most networks use the following commonly-defined Per-Hop Behaviors:

  • Default PHB—which is typically best-effort traffic
  • Expedited Forwarding (EF) PHB—dedicated to low-loss, low-latency traffic
  • Assured Forwarding (AF) PHB— which gives assurance of delivery under conditions
  • Class Selector PHBs—which are defined to maintain backward compatibility with the IP Precedence field.

[edit] Default PHB

A default PHB is the only required behavior. Essentially, any traffic that does not meet the requirements of any of the other defined classes is placed in the default PHB. Typically, the default PHB has best-effort forwarding characteristics. The recommended DSCP for the default PHB is '000000' (in binary).

[edit] Expedited Forwarding (EF) PHB

The IETF defines Expedited Forwarding behavior in RFC 3246. The EF PHB has the characteristics of low delay, low loss and low jitter. These characteristics are suitable for voice, video and other realtime services. EF traffic is often given strict priority queuing above all other traffic classes. Because an overload of EF traffic will cause queuing delays and affect the jitter and delay tolerances within the class, EF traffic is often strictly controlled through admission control, policing and other mechanisms. Typical networks will limit EF traffic to no more than 30%—and often much less—of the capacity of a link.

[edit] Assured Forwarding (AF) PHB

The IETF defines the Assured Forwarding behavior in RFC 2597. Assured forwarding allows the operator to provide assurance of delivery as long as the traffic does not exceed some subscribed rate. Traffic that exceeds the subscription rate faces a higher probability of being dropped if congestion occurs.

The AF behavior group defines four separate AF classes. Within each class, packets are given a drop precedence (high, medium or low). The combination of classes and drop precedence yields twelve separate DSCP encodings from AF11 through AF43 (see table)

Assured Forwarding (AF) Behavior Group
Class 1 Class 2 Class 3 Class 4
Low Drop AF11 AF21 AF31 AF41
Med Drop AF12 AF22 AF32 AF42
High Drop AF13 AF23 AF33 AF43

Some measure of priority and proportional fairness is defined between traffic in different classes. Should congestion occur between classes, the traffic in the higher class is given priority. Rather than using strict priority queueing, more balanced queue servicing algorithms such as fair queueing or weighted fair queuing are likely to be used. If congestion occurs within a class, the packets with the higher drop precedence are discarded first. To prevent issues associated with tail drop, the random early detection (RED) or weighted random early detection (WRED) algorithms are often used to drop packets.

Usually, traffic policing is required to encode drop precedence. Typically, all traffic assigned to a class is initially given a low drop precedence. As the traffic rate exceeds subscription thresholds, the policer will increase the drop precedence of packets that exceed the threshold.

[edit] Class Selector PHB

Prior to DiffServ, IP networks could use the Precedence field in the Type of Service (TOS) byte of the IP header to mark priority traffic. The TOS byte and IP precedence was not widely used. The IETF agreed to reuse the TOS byte as the DS field for DiffServ networks. In order to maintain backward compatibility with network devices that still use the Precedence field, DiffServ defines the Class Selector PHB.

The Class Selector codepoints are of the form 'xxx000'. The first three bits are the IP precedence bits. Each IP precedence value can be mapped into a DiffServ class. If a packet is received from a non-DiffServ aware router that used IP precedence markings, the DiffServ router can still understand the encoding as a Class Selector codepoint.

[edit] Advantages of DiffServ

One advantage of DiffServ is that all the policing and classifying is done at the boundaries between DiffServ clouds. This means that in the core of the Internet, routers can get on with doing the job of routing, and not care about the complexities of collecting payment or enforcing agreements.That is, DiffServ requires no advance setup, no reservation, and no time-consuming end-to-end negotiation for each flow, as with integrated services. This leads DS to be relatively easy to implement.

[edit] Disadvantages of DiffServ

[edit] End-to-end and peering problems

One disadvantage is that the details of how individual routers deal with the type of service field is somewhat arbitrary, and it is difficult to predict end-to-end behaviour. This is complicated further if a packet crosses two or more DiffServ clouds before reaching its destination.

From a commercial viewpoint, this is a major flaw, as it means that it is impossible to sell different classes of end-to-end connectivity to end users, as one provider's Gold packet may be another's Bronze. Internet operators could fix this, by enforcing standardised policies across networks, but are not keen on adding new levels of complexity to their already complex peering agreements. One of the reasons for this is set out below.

Diffserv operation only works if the boundary hosts honour the policy agreed upon. However, this assumption is naive[citation needed] as human beings rarely agree. A host can always tag its own traffic with a higher precedence, even though the traffic doesn't qualify to be handled with that importance. This in fact has already been exploited: Microsoft Windows 2000 always tags its traffic with IP precedence 5[citation needed], making the traffic classing useless. On the other hand, the network is usually quite within its rights to traffic shape and otherwise ration the amount of network traffic ingress with any particular precedence[citation needed], and so where this is enforced, overall network traffic flow provided to a host could be reduced by such a tactic.

[edit] DiffServ vs. More Capacity

Some people believe that the problem addressed by DiffServ should not exist, and instead the capacity of Internet links should be chosen large enough to prevent packet loss altogether.

The logic is as follows. Since DiffServ is simply a mechanism for deciding which packets to deliver or route at the expense of others in a situation where there is not enough network capacity, consider that when DiffServ is working by dropping packets selectively, traffic on the link in question must already be very close to saturation. Any further increase in traffic will result in Bronze services being taken out altogether. This will happen on a regular basis if the average traffic on a link is near the limit at which DiffServ becomes needed.

For a few years after the tech wreck of 2001, there was a glut of fibre capacity in most parts of the telecoms market, with it being far easier and cheaper to add more capacity than to employ elaborate DiffServ policies as a way of increasing customer satisfaction. This is what is generally done in the core of the Internet, which is generally fast and dumb with "fat pipes" connecting its routers.

However, this logic is flawed in many respects:

First, the problem of Bronze traffic being starved can be avoided if the network is provisioned to provide a minimum Bronze bandwidth, by limiting the maximum amount of higher priority traffic admitted.

Simple over-provisioning is an inefficient solution, since Internet traffic is highly bursty. Flash crowds occur, such as when Barack Obama was inaugurated as US president. If the network is dimensioned to carry all traffic at such times, then it will cost an order of magnitude more than a network dimensioned to carry typical traffic, with traffic management used to prevent collapse during such peaks.

It is not even possible to dimension for "peak load". In particular, when sending a large file, the TCP protocol continues to request more bandwidth as the loss rate decreases, and so it is simply not possible to dimension links to avoid end-to-end loss altogether: increasing the capacity of one link eventually causes loss to occur on a different link.

Finally, with wireless links such as EV-DO, where the air-interface bandwidth is several orders of magnitude less than the backhaul, QoS is being used to efficiently deliver VoIP packets where it would not otherwise be achievable.

[edit] Effects of dropped packets

Dropping packets wastes the resources that have already been expended in carrying these packets so far through the network. In many cases, this traffic will be re-transmitted, causing further bandwidth consumption at the congestion point and elsewhere in the network.[citation needed] To minimize this waste, packets must be discarded as close to the edge of the network as possible, while Diffserv is often implemented throughout a network (edge and core).[citation needed]

Thus, dropping packets amounts to betting that congestion will have resolved by the time the packets are re-sent, or that (if the dropped packets are TCP datagrams) TCP will throttle back transmission rates at the sources to reduce congestion in the network. The TCP congestion avoidance algorithms are subject to a phenomenon called TCP global synchronization unless special approaches (such as Random early detection) are taken when dropping TCP packets. In Global Synchronization, all TCP streams tend to build up their transmission rates together, reach the peak throughput of the network, and all crash together to a lower rate as packets are dropped, only to repeat the process.

Delays caused by re-scheduling packets due to Diffserv can cause packets to drop by the IPsec anti-replay mechanism.[citation needed]

[edit] DiffServ as rationing

Hence, DiffServ is for most ISPs mainly a way of rationing customer network utilisation to allow greater overbooking of their capacity. A good example of this is the use of DiffServ tools to suppress or control peer-to-peer traffic, because of its ability to saturate customer links indefinitely, disrupting the ISP's business model which relies on 1%-10% link utilization for most online customers.

[edit] Bandwidth Broker

RFC 2638 from IETF defines the entity of the Bandwidth Broker in the framework of DiffServ. According to RFC 2638, a Bandwidth Broker is an agent that has some knowledge of an organization's priorities and policies and allocates bandwidth with respect to those policies. In order to achieve an end-to-end allocation of resources across separate domains, the Bandwidth Broker managing a domain will have to communicate with its adjacent peers, which allows end-to-end services to be constructed out of purely bilateral agreements. Bandwidth Brokers can be configured with organizational policies, keep track of the current allocation of marked traffic, and interpret new requests to mark traffic in light of the policies and current allocation. Bandwidth Brokers only need to establish relationships of limited trust with their peers in adjacent domains, unlike schemes that require the setting of flow specifications in routers throughout an end-to-end path. In practical technical terms, the Bandwidth Broker architecture makes it possible to keep state on an administrative domain basis, rather than at every router and the DiffServ architecture makes it possible to confine per flow state to just the leaf routers.

[edit] DiffServ RFCs

  • RFC 2474—Definition of the Differentiated Services Field (DS Field) in the IPv4 and IPv6 Headers
  • RFC 2475—An Architecture for Differentiated Services
  • RFC 2597—Assured Forwarding PHB Group
  • RFC 3140—Per Hop Behavior Identification Codes (Obsoletes RFC 2836)
  • RFC 3246—An Expedited Forwarding PHB (Obsoletes RFC 2598)
  • RFC 4594—Configuration Guidelines for DiffServ Service Classes

[edit] References

  • "Deploying IP and MPLS QoS for Multiservice Networks: Theory and Practice" by John Evans, Clarence Filsfils (Morgan Kaufmann, 2007, ISBN 0-12-370549-5)
  • "Differentiated Services for the Internet", by Kalevi Kilkki, Macmillan Technical Publishing, Indianapolis, IN, USA, June 1999, is available in pdf-format at [1]

[edit] See also

[edit] External links

Personal tools