×

SUBSCRIBE TO TMCnet
TMCnet - World's Largest Communications and Technology Community

CHANNEL BY TOPICS


QUICK LINKS




 
NGN Magazine Magazine logo
Jan/Feb 2010 | Volume 2/Number 1
Feature Story

Traffic Management for Emerging Networks

By Sindhu Xirasagar

New applications and services are being deployed and accessed over communications networks at an unprecedented pace.

Today, tens of millions of wireline broadband subscribers regularly use telephony, IPTV services and the Internet. These applications and services are used for data exchange applications, such as email and file-sharing, video on demand, video collaboration applications, social networking and online gaming. Subscribers are increasingly demanding any-to-any communication – any service, on any device, at any time, anywhere. Long-term evolution standards are redefining wireless networks to meet these needs.

Significant differences in bandwidth, jitter and latency characterize various applications and services. While telephony applications consume relatively low sustained bandwidth (a typical voice connection is under 100kbps), they have stringent jitter and latency requirements. Video requires a high level of sustained bandwidth per channel (a few hundred kbps to several mbps), however some latency is tolerable. Online gaming and financial applications typically require relatively small amounts of extremely time-sensitive data. Applications accessed over the Internet have widely varying bandwidth requirements. Social networking applications, such as Facebook, can generate low to high-bandwidth sustained data flows that vary in duration from short bursts to longer sustained streaming. An added dimension to network bandwidth usage comes with file sharing applications. These types of applications consume significant upstream bandwidth due to the uploading of large files (e.g., videos).





Additionally, quality expectations for different applications and services vary and may even be specific to subscriber preferences. For ubiquitous, legacy applications, such as telephony, customers expect de facto good quality of service.

Customers also expect uninterrupted, cable TV-quality for IPTV channels. Best-effort is typically acceptable for Internet-based downloads and file-sharing applications (e.g., YouTube). For some services, such as online gaming and financial applications, some customers are prepared to pay additional fees for guaranteed service. The challenge for service providers is to maximize revenues and profits in this context of diverse requirements of applications, services and customer expectations. Consistently meeting the jitter and latency requirements of the various services is critical for customer retention. QoS for the various applications and services must match customer expectations, which as described above, could be variable.

By offering tiered service models, service providers can extract additional revenues from those customers willing to pay a premium for better QoS for specific applications and services. To cost-effectively support tiered service models, networks must be powered by sophisticated, fine-grained traffic management and statistics gathering capabilities. Today, tens of millions of wireline broadband subscribers regularly use telephony, IPTV services and the Internet. These applications and services are used for data exchange applications, such as email and file-sharing, video on demand, video collaboration applications, social networking and online gaming. Subscribers are increasingly demanding any-to-any communication – any service, on any device, at any time, anywhere. Long-term evolution standards are redefining wireless networks to meet these needs.

Significant differences in bandwidth, jitter and latency characterize various applications and services. While telephony applications consume relatively low sustained bandwidth (a typical voice connection is under 100kbps), they have stringent jitter and latency requirements. Video requires a high level of sustained bandwidth per channel (a few hundred kbps to several mbps), however some latency is tolerable. Online gaming and financial applications typically require relatively small amounts of extremely time-sensitive data. Applications accessed over the Internet have widely varying bandwidth requirements. Social networking applications, such as Facebook, can generate low to high-bandwidth sustained data flows that vary in duration from short bursts to longer sustained streaming. An added dimension to network bandwidth usage comes with file sharing applications.

These types of applications consume significant upstream bandwidth due to the uploading of large files (e.g., videos).

Additionally, quality expectations for different applications and services vary and may even be specific to subscriber preferences. For ubiquitous, legacy applications, such as telephony, customers expect de facto good quality of service.

Customers also expect uninterrupted, cable TV-quality for IPTV channels. Best-effort is typically acceptable for Internet-based downloads and file-sharing applications (e.g., YouTube). For some services, such as online gaming and financial applications, some customers are prepared to pay additional fees for guaranteed service.

The challenge for service providers is to maximize revenues and profits in this context of diverse requirements of applications, services and customer expectations. Consistently meeting the jitter and latency requirements of the various services is critical for customer retention. QoS for the various applications and services must match customer expectations, which as described above, could be variable.

By offering tiered service models, service providers can extract additional revenues from those customers willing to pay a premium for better QoS for specific applications and services. To cost-effectively support tiered service models, networks must be powered by sophisticated, fine-grained traffic management and statistics gathering capabilities.

Flow Isolation
Flow isolation is a key ingredient in networks supporting tiered service models. Flow isolation means separating different types of traffic destined for different customers used for different purposes into distinct flows for proper service level agreement enforcement. Network elements, especially in the access portion of the network, must be able to isolate flows at the service level (IPTV channel vs. VoIP channel). They must also be able to isolate flows at a customer level (file being uploaded vs. TV channel change command issued by the customer), and at an application level (file being uploaded or downloaded on YouTube vs. for data traffic access).

One or more flows may be assigned to a single queue. Different traffic management schemes and priorities may then be applied to traffic in different queues based on service-specific requirements and/or customer-specific SLAs. Flow isolation also is essential for tracking statistics used for billing and accounting purposes and is especially important to implement tiered service models.

Traffic Management
Traffic management performs three important functions in a network designed to meet service-specific requirements modulated with subscriber-specific SLAs including:
• checking conformance (i.e., policing);
• making discard decisions (i.e., buffer management);
• and regulating traffic flow in various queues (i.e., traffic scheduling).

Policing Policing algorithms check whether a given subscriber is consuming bandwidth beyond that allowed by the associated SLA. The Two Rate Three Color Marker scheme described in IETF RFC 2698 is an example. It consists of metering of incoming subscriber traffic packets and marking them as red, yellow or green. The packets are marked based on parameter values for the two rates, namely peak information rate and committed information rate and their associated burst sizes. By assigning different values to the parameters for different subscribers, traffic from any given subscriber can be effectively policed for a specific SLA.

Buffer Management A buffer management algorithm, such as random early discard or weighted random early discard, can be used to support a tiered service model. In WRED, the ability to define depths, drop thresholds and weights per queue enables the definition of multiple service classes. For example, gold, silver and bronze can be associated with different subscriber SLAs. Drop policies for red, yellow and green packets as a result of policing can be specified. The algorithm drops packets randomly with a probability related to class of service and depth of the queue.

Traffic Shaping
Traffic shaping enables controlling traffic flow in different queues according to associated criteria. For example, higher priority packets are scheduled before lower priority packets. Often, configurable hardware schedulers are used to implement traffic shaping. For instance, a constant bit rate scheduler is characterized by very low jitter and is typically used for voice traffic. While configurable hardware schedulers are deterministic, they are inflexible and cannot adapt to evolving traffic conditions. Purely software-based traffic schedulers enable implementation of sophisticated traffic shaping algorithms that can be evolved over time, but lack the determinism of hardware schedulers.

Some network processors include dedicated programmable hardware accelerators to implement traffic schedulers. They combine the advantages of the determinism of hardware schedulers with the flexibility of software-based schedulers. With these devices it is possible to program a menu of traffic shaping algorithms that allow each service provider the ability to select and configure the desired algorithms tailored to specific services and SLA offerings. Equally important, these traffic-shaping algorithms can be evolved over time to adapt to changing network conditions and traffic profiles, without expensive hardware replacements.

Traffic shaping algorithms continue to evolve to meet changing requirements. Earlier DSL-based triple play (Internet, VoIP and IPTV) service deployments used weighted round robin for traffic scheduling. WRR, shown in Figure 1, enables prioritizing and assigning different weights to traffic in different queues. WRR was effective as long as Internet traffic mostly consisted of low-bandwidth, text-based Web access. With increasing adoption of P2P applications, high-bandwidth Internet traffic began impacting IPTV QoS. If the usage of different services for a given subscriber is pre-determined, it is technically possible to tune and pre-configure the algorithm parameters to guarantee QoS for voice and video channels. However, this is a complex task, and when usage patterns change and QoS suffers, re-tuning is required. The net effect is dissatisfied customers combined with high opex for special training and continual re-tuning.

A multi-level hierarchical scheduling algorithm depicted in Figure 2 enables service providers to isolate services and assign priorities across them. It also allows service providers to group like services and allocate guaranteed bandwidth to the high priority groups carrying jitter and/or latency sensitive traffic (e.g., voice, video). This ensures that such traffic is not disrupted by intermittent bandwidth-intensive Internet traffic such as that resulting from video-sharing applications (the left group in Figure 2). In addition, traffic parameters for all services can be pre-configured based on known bandwidth and service categories. For instance, the maximum bandwidth of each digital TV channel is 19mbps. If the maximum number of simultaneous channels being offered is eight, the total maximum bandwidth for IPTV is 152mbps. The video traffic group is allocated guaranteed bandwidth of 152mbps. Services and subscribers can be added without having the operator tune traffic management for each new subscriber. At the same time, service providers can create additional configurations with associated traffic management parameters. For example, a different configuration may be used to configure those subscribers using voice and data only (no IPTV). In this case, the remaining line bandwidth is allocated to the data traffic group, thus enabling higher fees for the higher-bandwidth data service. This dynamic configurability also presents the opportunity of creating flexible, on-the-fly SLAs.

When used in conjunction with fine-grain flow isolation, traffic management can be leveraged to implement advanced security measures. Increasingly, security threats need to be detected and thwarted deeper in the network. Such threats include denial of service and distributed DoS attacks. With security-enabled traffic management, packets belonging to various control protocols, such as TCP Syn-ACK, may be isolated into separate flows and each flow policed. The subsequent buffer management and traffic shaping algorithms can be configured for that flow in such a way that all packets marked red or yellow are dropped. Additionally, the rate is configured low enough to prevent the flood of control packets (constituting the attack) from reaching the subscriber. All of these actions combined prevent the DoS attack from bringing down the services and computers. However, to implement this capability, the underlying flow isolation, policing, buffer management, and traffic shaping engines must be programmable.

Emerging wireline and wireless networks must be built to service a rapidly growing number of applications and services, some of which haven’t even been envisioned yet. Effective traffic management is an important factor for cost-effective delivery of multi-play services with guaranteed QoS. Network processors with programmable flow isolation and traffic management engines are critical for enabling strategies that combine QoS and cost-effective bandwidth management. Such processors also enable service providers to adapt to changing traffic patterns without incurring the cost of expensive hardware upgrades.



Sindhu Xirasagar is the product line manager for software and system solutions for LSI Corp. (www.lsi.com).

NGN Magazine Table of Contents









Technology Marketing Corporation

2 Trap Falls Road Suite 106, Shelton, CT 06484 USA
Ph: +1-203-852-6800, 800-243-6002

General comments: [email protected].
Comments about this site: [email protected].

STAY CURRENT YOUR WAY

© 2023 Technology Marketing Corporation. All rights reserved | Privacy Policy