Traffic management for affordable broadband

Efficient traffic management is the key to delivering QoS and increasing provider revenue for personal broadband services, argues Agere's Systems Applications Engineering Manager, Deepak Kataria.

The advances in xDSL technology and its deployment by providers are making possible availability of high-speed, broadband access to an increasing base of subscribers.

These include residential customers and business customers.

With high-speed broadband access, not only is Internet access being delivered with increased speed but new multimedia services are also becoming available.

These services include broadcast video, audio/video conferencing, collaborative editing, video-on-demand (VoD), PPV (pay-per-view movies and content stored in the network and streamed on demand), Internet telephony, Internet radio, premium interactive content, and interactive gaming.

The delivery of such services opens new revenue streams for telecomms service providers.

Customers are willing to pay for such a rich suite of services, especially when multiple services are bundled and made affordable.

Businesses are also looking to cut communications costs by replacing dedicated networks with VPNs and use multiservice offerings.

Residential bundled services can include the "triple play" of voice, video and data services all delivered over the same shared copper by a single provider.

Business bundled services can include VPNs for interconnecting multiple corporate LANs and for secure telecommuter access, video conferencing, multiline Internet telephony, enhanced voice and messaging services, business video services also all delivered by a single provider.

Whereas best effort service has been traditionally used to provide Internet access service such as like web browsing, the new multimedia services require end-to-end quality-of-service (QoS) guarantees.

These QoS technology features are made possible using traffic functions on network processor chips.

Whereas interactive multimedia applications require bounded delay and bandwidth guarantees, high-speed data services and stored audio/video streaming services require bandwidth guarantees.

Besides QoS, support for additional control functions is needed to support the various on-demand services.

ISPs/portals and other content providers (network service providers) are connected to the network access providers, which connect via a broadband access network to the subscribers.

The xDSL links from the subscribers terminate in digital subscriber line access multiplexers (DSLAMs), which is part of the broadband access network.

Partnerships exist between network service providers and network access providers to deliver the various content driven broadband services to the subscribers.

The network access provider makes sure connectivity toward the network service provider is available, whereas each network service provider takes care of the services themselves.

From the perspective of QoS controls, traffic management in the DSLAMs can be particularly challenging because solutions have to be implemented at different levels of control granularity: at the level of the individual services active within a given subscriber; at the level of the individual xDSL link load for the given subscriber; at the level of aggregate subscribers supported on a given line card; and at the level of aggregate line cards supported on a given uplink card.

There may be other intervening levels of control granularity that have to do with the virtual partitioning of the links at the various levels of the hierarchy for better QoS controls.

For example, macro-level controls will be needed to differentiate residential customers from business customers.

Furthermore, DSLAMs themselves may be interconnected in a star or daisychain configuration through what is known as subtending.

In either implementation, multiple systems connect to one system that aggregates transmission of all attached systems and provides one network uplink.

The aggregating system is known as the subtending system; each of the systems connected to it are known as subtended systems.

This many-to-one relationship at different levels is characteristic of DSLAMs architectures.

This one-way direction of aggregation from the subscribers toward the provider is also known as the upstream direction.

The opposite direction from the provider to the subscribers is known as the downstream direction.

Despite the aggregation in the upstream direction, the subscribers' traffic needs to remain segregated.

This can be achieved by providing an equivalent mirrored structure of controls in the downstream direction.

The overall objective of traffic management is to support a wide variety of services with diverse QoS requirements while ensuring efficient sharing of network resources.

The former allows providers to offer new revenue generating services; the latter promotes service affordability.

Traffic management includes functions such as policing, buffer management, scheduling, shaping, backpressure flow control, CAC admission control, route selection, and node/network monitoring.

These functions must support full bandwidth flexibility such that upstream and downstream rates can be chosen freely and continuously up to the maximum physical limits.

The functions also should enable full service flexibility such that a random mix of services with various bit rates and various traffic requirements can be supported, within available bit rate limits.

The functions should foster maximum sharing of network resources and help minimise network downtime.

These functions can be implemented in a centralised manner where all the control functionality resides in an uplink card that handles traffic from all the connected line cards, as well as all the subscriber traffic from each line card.

In the distributed approach, the control functions are distributed between the uplink card and the line card.

The distributed approach provides for a more scalable and flexible architecture.

In the upstream direction, policing ensures traffic conforming to the SLAs is allowed into the network.

Marking of traffic may be employed to signal the relative treatment of traffic in the network.

Differentiated traffic scheduling is needed to treat the delay-sensitive real-time services (telephony or traffic with real-time video content), preferentially over the data services.

Class-specific backpressure flow control is needed to handle the periods of temporal overload, when the incident traffic load exceeds the available bandwidth in the upstream direction.

Due to the fan-out topology, no traffic aggregation exists in the downstream direction.

Rate shaping is employed so downstream traffic flows smoothly with no likelihood of buffer overflows.

Temporary rate excesses are accommodated by limited downstream buffering.

Traffic multicasting is needed at the edge node such that the WAN links upstream can be efficiently used.

Priority-based traffic handling is needed so low latency can be provided for high-quality telephony, audio and video conferencing and other types of video services with real-time content.

The call admission control (CAC) function keeps track of upstream and downstream bandwidth along all the nodes and links in the path.

Any new connection requesting a certain QoS level is only admitted if the resulting bandwidth use is within a predefined admissible region.

Even though the service discipline is based on strict priorities, the buffer analysis for the top three QoS classes is identical.

This is because each QoS class can be flow-controlled independently.

It is possible for the flow control flag from the upstream node to be turned off for class 2 traffic, while the flow control flag for class 1 is turned on.

Let R be the datarate in bits per second on each trunk and f be the period (duration) of the flow control feedback to the downstream device.

Latency in the flow control includes the round-trip propagation delay from the upstream device to the downstream device as well as the processing delays at each device.

Let h be the aggregate latency.

If there are N links at rate R feeding into a queue, the maximum rate at which data can arrive at the queue is NR bit/s.

Clearly, the headroom should be at least NR(f+h) bit.

Flow control is applied at the packet level; that is, when a device receives a flow control flag from the upstream node, the packet currently being transmitted is completed.

Then further transmission is turned off.

Thus, the headroom should be further augmented by one maximum-sized packet per link.

If the maximum packet size is M byte (8M bit), then for each of the CBR, VBR-rt and VBR-nrt queues, the headroom is NR(f+h) + N8M bit.

In the case of UBR, there is one queue per incoming link, so for each UBR queue, headroom (UBR) is R(f+h) + 8M bit.

For downstream traffic, the CAC function needs only ensure aggregate bandwidth use for guaranteed traffic remains within the available capacity.

For upstream traffic, the admissible region is the set of bandwidth use values for which all guaranteed flows receive their contracted QoS.

Network/node monitoring functions allow the detection of resource problems proactively such that remedial actions can be taken for maximum availability.

For example, switch level redundancy at the line card; and switch fabric, common control, and power supply levels and network level redundancy, such as automatic routing, failure recovery, and congestion controls are important functions that help minimise network downtime.

Traffic management can thus be seen as playing a critical role to support new revenue generating services for service providers.

Efficient traffic management also plays a major role in maximising use of network resources.

By using the minimum amount of buffer space for the maximum good throughput; by satisfying the QoS requirements; by using the least amount of bandwidth; by supporting the maximum number of connections over a given network link; and by minimising network downtime, traffic management helps increase provider revenue.

This makes personal broadband services affordable.

Efficient traffic management facilitates logarithmic increase in cost of service with bandwidth demand.

This occurs by deriving maximum possible use from existing and newly deployed network and system resources.

In the absence of this, existing and new capital investments are underused, leading to the cost of service growing linearly with bandwidth demand.

Efficient traffic management also offers several hooks for increasing provider revenue.

By dynamically adjusting the scheduling weights for both improving QoS for higher priority users-those that suffer from a high population of users with lower priority - revenue earned by the service provider can be increased.

However, this strategy should not completely block the users who may have subscribed for cheaper service.

Users can also be charged additional fees depending on the throughput improvement, for example, when real-time services are not being used.

Traffic management functions in this case must detect the change and provide the new rate to the willing subscriber.

Thus, efficient traffic management plays a major role in delivering QoS and increasing provider revenue, both of which are key elements in making affordable personal broadband a reality.

Back to top