QoS Protection

Carriers are in the midst of designing next-generation, IP-based wireless networks structured on new and faster radio-access technologies including 2.5G, 3G and wireless LANs. These access networks combined with new packet-based mobile core networks can allow carriers to deliver high-value, differentiated, IP-based applications and services cost-effectively, as long as there are effective QoS mechanisms in place. Therein lies the challenge.

Mobile environments raise several hurdles to QoS. Even though the new mobile access networks are more efficient than today's 2G circuit-based environments, they still don't offer enough bandwidth to bring high levels of service to all customers. In addition, like all wireless services, the access network has highly variable quality levels. As a result, carriers must provide customers fair access to the network's limited bandwidth, and they must optimize transport for IP applications to deliver data seamlessly between the mobile and fixed Internet. For mobile Internet carriers to provide high-quality and differentiated wireless IP services, QoS mechanisms must be purpose-built within the infrastructure equipment designed to merge these networks.

The wired IP world and the mobile wireless world are two fundamentally different networks, and they require different strategies for providing QoS. Somehow, an end-to-end IP network that spans the wired and wireless worlds must reconcile these differences.

Setting Traffic Priorities

In the IP world, traffic patterns are random with frequent data bursts, and QoS mechanisms deal mostly with bandwidth availability and prioritized treatment of traffic. Although ideally deployed end to end, QoS is most important in network segments that encounter the most congestion due to limited available frequency. In network segments that do offer QoS, the most common method for ensuring it is differentiated services (DiffServ).

DiffServ is a Layer 3 protocol that controls IP traffic flow based on identifying which types of traffic should receive precedence. Forwarding decisions can be made on a packet-by-packet level according to policies established by the carrier. For example, high-priority, low-latency packets are sent to a higher-priority queue and transmitted more frequently than lower-priority background traffic. (Service-level fulfillment requires proper traffic engineering because DiffServ does nothing to reserve bandwidth or resources.) DiffServ is unique in that it allows a simple migration of QoS into an end-to-end resource that may not be controlled by a single carrier or where the carrier may have an end-to-end resource that is not uniformly QoS enabled.

As the name suggests, end-to-end QoS requires that carriers guarantee service levels at the access points and in the backbone alike. However, it is much more difficult to bring QoS to the access pipes because they are more congested. To provide end-to-end QoS, carriers must ensure that access traffic is not affected by the congestion.

Carriers can use DiffServ queuing mechanisms to allocate resources fairly. Schemes such as random early detection offer fair congestion management because they inherently perform fair management of best-effort traffic. Carriers also can mark data according to the service-level agreement (SLA) to ensure traffic exceeding the service-level guarantee can be discarded in the event of congestion.

However, setting traffic priorities using DiffServ is not enough to provide end-to-end QoS, and neither is bandwidth availability in the core. No matter how fat the pipelines in downstream IP backbone networks, access networks and air-link quality ultimately will affect transfer rates. In addition, the fixed IP network and the mobile core use different methods to track subscriber QoS policies.

In the mobile core, network devices know the priority status of data packets. These priorities must be mapped into the fixed IP network so that the intent of the QoS policy will be maintained from end to end.

Bridging Dissimilar Networks

Wireless environments inherently have greater error rates, higher delays and more limited access resources than fixed networks. Many of the techniques used to minimize errors within these networks contribute to additional end-to-end network delay. Although not all applications are sensitive to these factors, they must optimize transmission control protocol-based transport effectively to deliver high-quality performance over 2.5G and 3G networking environments.

To provide end-to-end QoS that bridges the wireless and terrestrial IP networks, each element along the data path must conform to a traffic contract, as shown in Figure 1. Moreover, carriers must map IP-based network service classes (or specific IP applications) to wireless network QoS service classes to ensure QoS is provided through access networks.

To implement these traffic contracts, carriers can leverage QoS protocols and standards that affect the quality of traffic flow between network elements and that dictate how to handle congestion within a network element. QoS mechanisms especially are needed for network congestion points.

Implementing QoS

Any network device (such as a router, air link or gateway) and data link can be a congestion point for IP traffic flows and should contain a QoS mechanism. To provide fair service, the network device must provide queuing so that resources are allocated appropriately among different traffic streams. If congestion occurs, the device also must treat higher-priority traffic better than lower-priority traffic based on metrics defined by the carrier. The latest packet-switched radio access architectures including GPRS and UMTS define QoS profiles that are negotiated between a mobile station and a serving GPRS support node (SGSN). Built on top of the GSM network architecture, GPRS and UMTS provide a common packet-switched core network designed to support multiple levels of QoS.

Carriers can offer service classes based on a variety of QoS classes. The QoS negotiation for both GPRS and UMTS uses a QoS profile that defines the expected QoS based on characteristics including precedence, delay, reliability, peak throughput and mean-throughput classes.

GSM/TDMA timeslot scheduling over the radio link (a requirement for tiered services) is complex, so carriers are unlikely to support GPRS-specified QoS profiles during the next two years. Once this issue is resolved, however, carriers most likely will begin to offer two service levels priority and regular service.

In contrast, UMTS using WCDMA does not have to address prioritized scheduling of users across timeslots on the air interface. Rather, UMTS supports four QoS classes built from appropriate QoS profile definitions. The gateway GPRS serving/support node (GGSN) maps these classes to the appropriate IP-based QoS class. As IP is pushed further into the mobile access network, IP-based mechanisms can be used further into the network to deliver fair and priority access as traffic is transmitted through the thin terrestrial links between the SGSN and the mobile terminals.

Billing & Peering Challenges

To make advanced services pay, carriers that rely on QoS must fully define billing models. However, because the radio interface and/or mobile-access network infrastructure may downgrade a QoS request, carriers cannot absolutely guarantee levels of service. As a result, they must establish tariffs to support downgraded service.

Carriers likely will provide QoS services within their own networks first to facilitate access to carrier-provided content and services. To offer guaranteed service levels for real-time services such as voice and multimedia, however, the mobile Internet infrastructure also must provide predictable QoS behavior between providers including those using best-effort services on public links such as the Internet. Nevertheless, even without peering relationships between providers, operators can provide low- and high-priority services by deploying QoS within resource-constrained access networks and by paying attention to congestion in application servers.

QoS should be viewed as a set of capabilities that enables effective support for IP applications, provides fair access to limited network resources and allows carriers to offer premium SLAs all factors in increasing the profitability of the mobile Internet.

---

Sangillo (msangillo@megisto.com) is Megisto Systems product manager.