Through effective use of QoS, service providers can deliver the tailored services that enterprise customers demand and differentiate themselves from competitors.
EXECUTIVE SUMMARY
Enterprises demand more customizable services from service providers, including higher-bandwidth offerings and more tailored bandwidth services, as well as support for multiservice traffic. Quality of Service (QoS) techniques, including classification and marking, traffic conditioning, congestion avoidance, and congestion management, help providers create service-driven networks to meet customer needs and enhance revenue. The comprehensive Cisco® Metro Ethernet portfolio encompasses a variety of technologies and product options, offering service providers an end-to-end approach to deploying profitable Metro Ethernet services.
In an increasingly competitive marketplace, service providers that can tailor services best will be more likely to attract the maximum market share and customer revenue. This capability is essential to meet the needs of customers that are deploying more advanced enterprise applications like storage, videoconferencing, and hosted IP telephony.
QoS helps service providers to differentiate themselves from competitors and to scale capacity more effectively. It can control traffic attributes such as delay, jitter, or packet loss to efficiently use network resources and help ensure consistent application performance during congestion events. In aMetro Ethernet environment, service providers can implement QoS using several techniques, and these techniques can be implemented at the user-facing provider edge (U-PE), aswell as at the network provider edge (N-PE).
Cisco Systems® offers a variety of Metro Ethernet service solutions. Each Cisco Metro Ethernet solution enables providers to support QoS services classes to meet specific customer needs. By applying effective QoS mechanisms, service providers can develop service-level agreements (SLAs) toalign their services closely with customer applications, and tap into a number of benefits, including increased revenue.
CHALLENGE: SERVICE-DRIVEN NETWORKS ON THE RISE
Enterprises continue to demand more customizable services from their service providers, including higher-bandwidth offerings and more tailored bandwidth services. They are seeking ways to provide support for data, voice, and video traffic according to customers' specific business objectives and applications. And they need ways to customize traffic differently, depending on the type of application. In a competitive marketplace, service providers that can tailor services most effectively will be more likely to attract the maximum market share and customer revenue.
To deliver the tailored services that customers demand, providers are changing their approach to designing and deploying networks. In the past, they focused on building a network infrastructure, then on delivering the services that the infrastructure could support. Today, services are increasingly driving the network infrastructure. This is especially true with Metro Ethernet networks.
To create new revenue, providers are expanding their service offerings with connectivity services like Layer 2 and Layer 3 VPNs, as well as value-added services like storage, videoconferencing, and hosted IP telephony-all enabled by Metro Ethernet networks. Service providers need solutions that provide the ability to deploy multiple services over a common Metro Ethernet infrastructure; this greatly improves their return on investment (ROI).
SOLUTION: ENABLING SERVICE LEVEL AGREEMENTS WITH QUALITY OF SERVICE
Consistent, End-to-End Quality of Service
QoS is an important technology consideration for creating service-driven networks, providing tighter bandwidth control and support for SLAs. QoSmechanisms can control traffic attributes to efficiently use network resources and help ensure consistent application performance during congestion events. And effective QoS helps service providers to differentiate themselves from competitors and to scale capacity more effectively.
In point-to-point connections, traffic attributes can be enforced at each network ingress point, and traffic profiles can easily be defined, implemented, controlled, and measured. Layer 3 devices, such as routers, can offer a rich set of QoS features. For multipoint-to-multipoint services, QoS definitions are more complex, and service providers need to enforce traffic contracts both at the network ingress and egress. QoS not only enables clearly defined SLAs, it also increases the transport efficiency of the network. Using intelligent packet processing together with QoS, service providers can oversubscribe their networks to make better use of their existing interfaces and bandwidth.
By applying the QoS mechanisms discussed above, service providers can develop SLAs to align their services closely with customer applications, and tap into a number of benefits including additional revenue without major architectural change.
Enhancing Revenue with Service-Level Agreements
Enabling SLA revenues improves Net Present Value (NPV) more than any other factor. NPV puts the value of future cash flows (revenue and expenses) in present-day values. By moving customers from basic to advanced service levels, service providers can collect greater revenues and improve the business case. In the sensitivity analysis, increasing revenues through SLAs by 10 percent raised the NPV by 13 percent.
Service providers offering Metro Ethernet access services can secure greater revenues from their clients in three ways. First, they can offer a variety of increasingly valuable and higher-priced SLAs. This approach requires differentiating service levels and pricing each to reflect the perceived value of the SLAs.
Secondly, service providers can deliver different classes of service that correspond to a variety of Committed Information Rates (CIRs). For instance, a service provider could offer three classes of service-bronze, silver, and gold-each offering higher performance than the previous class. An end user might put voice traffic on a gold class of service, backup traffic on a silver class, and basic, noncritical data on a bronze class. By varying characteristics such as prioritization, latency, jitter, and packet loss, service providers can differentiate the service classes and price them accordingly.
Third, service providers can offer dynamic bandwidth allocation or bandwidth on demand for applications such as video broadcasts or data backups. For example, an enterprise customer might want to broadcast a CEO's speech to different sites in real time. To accommodate the broadcast, the customer needs to increase its network bandwidth from 100 Mbps to 200 Mbps. The service provider would make the additional 100 Mbps available and charge for the increased bandwidth for the broadcast's duration.
Enabling SLAs permits the service provider to create layers of differentiated services. They increase value by capturing more revenue per customer, as well as expanding the addressable market given strict SLA requirements among certain customer segments.
Financial Benefits of Upgrading Service-Level Agreements
For example, assume that a customer has a 100-Mbps Metro Ethernet access service with a 55-percent CIR. When subscribing to a basic-level SLA, the customer generates average revenue of US$3165 a month. By moving the customer to a premium SLA service, the customer generates revenue of$3640 a month-a 15-percent increase. At the same time, the service provider's monthly costs rise from $1172 to only $1201, a 2.4-percent difference due to sales commission increase. The net margin consequently rises by $446 a month, increasing from 63 percent to 67 percent of revenue.
By persuading this customer to move from a basic SLA to a premium SLA, the service provider increases its revenue. By automating the move, the provider incurs no additional costs, other than a slightly higher sales commission. Because the service provider uses the same circuit and supports that circuit the same way, other costs remain unchanged.
IMPLEMENT QUALITY OF SERVICE USING FOUR DIFFERENT TECHNIQUES
To deliver differentiated service, service provider networks deliver a particular kind of service based on the QoS specified by each packet or frame (Figure 1). This specification can occur in different ways, including using the IP Precedence bit settings in IP packets or the 802.1p Class of Service (CoS) bits in the Ethernet header. The network uses the QoS specification to classify, mark, shape, and police traffic and to perform intelligent queuing. In a Metro Ethernet environment, service providers can implement QoS using several techniques:
• Classification and marking
• Traffic conditioning
• Congestion avoidance
• Congestion management
Figure 1. QoS Functions for Metro Ethernet
Classification and Marking
Classification categorizes network traffic into predefined classes. These classes are essentially "buckets" that map to specific traffic properties, such as priority and latency.
Classification
Traffic is normally classified as it enters the network, where it is marked for appropriate treatment. After the traffic has been classified and marked atthe edge of the network, the network must be set up to provide differential service to the various traffic flows. Common methods to differentiate traffic include Layer 2 CoS or 802.1p, Layer 3 Type of Service (ToS), or Layer 3 Differentiated Services Code Point (DSCP).
Choosing the appropriate method to differentiate traffic is an essential part of building a QoS model. Providers should carefully consider the network inputs (traffic streams) and the overall system consistency, especially when interfacing to switches and routers.
Marking
The Class-Based Packet Marking feature lets providers efficiently mark packets, and then differentiate them based on these markings. This occurs atthe Ingress of the network. The Class-Based Packet Marking feature allows service providers do the following:
• Mark packets by setting the IP Precedence bits or the IP DSCP in the IP ToS byte
• Mark packets by setting the Layer 2 CoS value
• Mark packets or frames by setting the Multiprotocol Label Switching (MPLS) Experimental bits
• Associate a local QoS group value with a packet; class-based marking lets providers create traffic classes that are given specific treatment at each QoS domain
Traffic Conditioning
Service providers can employ two kinds of traffic-conditioning mechanisms for QoS: policing and shaping. Providers can deploy these features throughout the network to help ensure that a packet or data source adheres to a stipulated contract, and to determine the QoS to render to the packet. Both policing and shaping mechanisms use the traffic descriptor for a packet, indicated by the classification of the packet, to help ensure adherence and service.
Policing
A policer drops traffic that is out of profile. For example, the rate-limiting policer either drops the packet or marks down the CoS, resetting the ToS bits in the packet header.
Policers can have two or three colors. A two-color policer lets providers identify two types of traffic: conforming traffic that meets the CIR, and exceeding traffic that goes above the CIR.
A three-color policer lets providers identify three types of traffic: conforming traffic, exceeding traffic that meets the Peak Information Rate (PIR), and violating traffic that goes above the PIR.
Shaping
A shaper delays excess traffic using a buffer or queuing mechanism to hold packets and shape the flow when the data rate of the source is higher than expected. Traffic shaping and policing can work in tandem. For example, a good traffic-shaping scheme can make it easy for nodes inside the network to detect misbehaving flows. Shaping is not optimal for real-time traffic, because it introduces additional delays.
As with Frame Relay environments, Metro Ethernet customer premises equipment (CPE) must perform traffic shaping whenever possible on the Ethernet WAN links to minimize policer-based drops. A high number of these drops might adversely affect the performance of TCP-based applications.
Congestion Avoidance
Congestion-avoidance techniques monitor network-traffic loads to anticipate and avoid congestion at common network bottlenecks. Congestion avoidance starts dropping packets when they reach the congestion level. One commonly used congestion-avoidance mechanism is Random Early Detection (RED), which takes advantage of TCP's congestion-control functions. RED is best for high-speed transit networks.
Another congestion-avoidance technique, Weighted RED (WRED), drops packets selectively based on traffic marking. Packets with a higher IP precedence are less likely to be dropped than packets with a lower precedence.
Congestion Management
Congestion-management features control congestion when it occurs. One way that network elements manage an overflow of arriving traffic is to use a queuing algorithm to sort the traffic, and then determine some method of prioritizing it onto an output link. Each queuing algorithm was designed to solve a specific network-traffic problem and has a particular effect on network performance.
Some examples of congestion management include the following:
• Priority Queuing-This allows providers to define how traffic is prioritized in the network. Providers can define a series of filters based on packet characteristics to cause the router to place traffic into queues.
• Class-Based Weighted Fair Queuing (CBWFQ)-This lets providers support user-defined traffic classes. They can define traffic classes based on match criteria including protocols, access control lists (ACLs), and input interfaces.
• Weighted Round Robin (WRR)-WRR supports flows with significantly different bandwidth requirements. Each queue can be assigned a specific percentage of bandwidth.
• Weighted Deficit Round Robin (WDRR)-WDRR lets providers support weighted, fair distribution of bandwidth for queues that contain variable-length packets.
QUALITY OF SERVICE IN YOUR METRO ARCHITECTURE
Each of the QoS techniques discussed thus far comes into play at a specific point in the network architecture. QoS roles can be assigned at the user-facing provider edge (U-PE), as well as at the network provider edge (N-PE).
Quality of Service at the User-Facing Provider Edge
At the edge of the network, the U-PE plays an important role in terms of managing major QoS functions (Figure 2). On ingress, classification and marking of the traffic are coupled with policing to ensure that traffic is being treated according to the service-class definition. On egress, queuing and scheduling of the traffic classes provide congestion management.
Figure 2. QoS Functions at the U-PE
Quality of Service at the Network Provider Edge
Ingress QoS functions (Figure 3) at the N-PE include copying the packet 802.1p bits into the MPLS Experimental bits (EXP bits). Egress functions include queuing and scheduling to help ensure proper congestion management.
Figure 3. QoS Functions at the N-PE
METRO ETHERNET QUALITY OF SERVICE DOMAINS AND CLASSES
To support the Ethernet connectivity and Ethernet access-based value-added services, Cisco offers a variety of Metro Ethernet service solutions (see Toward a Service-Driven Metro Network for additional details on these solutions). Each Cisco Metro Ethernet solution helps providers to support QoS services classes to meet specific customer needs.
End-to-end Metro Ethernet service involves multiple QoS domains. The domains are described in Figure 4, which shows the elements that make up a Metro Ethernet architecture and their corresponding QoS domains.
Figure 4.
Metro Ethernet QoS Domains
Metro Ethernet QoS domains include the following:
• Enterprise network-The first domain is the enterprise network where DSCP is typically used to mark and identify traffic classes. There are up to64 different classes of services.
• Service provider Ethernet access domain-The access network, from U-PE to N-PE, is a Layer 2 Ethernet domain. Here, the provider uses Ethernet 802.1p bits to mark customer traffic. At the U-PE, the provider performs QoS traffic classification, marking, and admission control.
• Service provider MPLS core-From the values marked at the Ethernet access domain, the Ethernet 802.1p bits are copied into MPLS experimental bits of the packets traveling in the MPLS core network.
DIFFERENT QOS MARKINGS FOR DIFFERENT METRO APPLICATIONS AND SLA LEVELS
When traffic enters the service provider network, it is classified and marked according to the service class offered. For example, service providers will mark most data service such as Internet browsing as Best Effort and use the lowest SLA. But for the applications that have higher jitter, delay, orpacket-loss requirements, service providers will mark traffic differently and give it higher priority. Table 1 shows an example of how the classes are marked in the Ethernet Access Domain.
Table 1. Ethernet Access Domain Classes
Application
Layer 2 CoS Value
Service Provider Network Management Traffic
7
IP Telephony Voice Traffic
5
Interactive Video Traffic
4
IP Telephony Call Setup and Teardown
5 or 3
Business-Critical and Streaming Video
2
Best Effort
0
COMPLETE SOLUTIONS FOR METRO ETHERNET NETWORKS
The comprehensive Cisco Metro Ethernet portfolio encompasses a variety of technologies and product options, offering service providers an end-to-end approach to deploying profitable Metro Ethernet services. The portfolio includes the following:
• IP/MPLS and Ethernet
– Cisco 12000 Series Routers
– Cisco 10700 Series Routers
– Cisco 7600 Series Routers
– Cisco Catalyst® 6500 Series Switches
– Cisco Catalyst 4500 Series Switches
– Cisco Catalyst 3750 metro Series Switches
– Cisco Catalyst 3550 Series Switches
– Cisco Catalyst 2950 Series Switches
– Cisco Catalyst 2950 LRE Series
• SONET/SDH and Ethernet
– Cisco ONS 15600 Multiservice Switching Platform (MSSP)
– Cisco ONS 15454 SONET/SDH Multiservice Provisioning Platform (MSPP)
– Cisco ONS 15300 Series
• Dense Wavelength-Division Multiplexing and Coarse Wavelength Division Multiplexing
– Cisco ONS 15454 Multiservice Transport Platform (MSTP)
