Unified Policy Management for Uniform Service Delivery in an IMS Network

By Matt Tooley

One of the many promises of IP Multimedia Subsystem (IMS) is its ability to offer managed IP services enhanced with premium delivery. The IMS architecture addresses management of quality of service (QoS) for these services through the Policy Decision Function (PDF), which is responsible for dynamically allocating network quality of service (priority, bandwidth, jitter, latency) in order to ensure the network performance of managed services for individual subscribers.Originally envisioned as part of the Next Generation Network (NGN) architecture for mobile broadband, due to its adoption in mobile, IMS has crossed over into other broadband access technologies. IMS, or permutations of it, are being developed by nearly all the applicable telecommunication standards bodies for cable, DSL, WiMAX, and 3G to meet service providers’ strategy of providing fixed/mobile convergence (FMC) services to
their customers.

With IMS’ adoption as the architecture of choice for FMC, operators face a unique set of new challenges with regard to policy management: they must ensure that applications and services being deployed in the context of the IMS framework provide a consistent quality of experience across the heterogeneous multi-network infrastructure of FMC. This article discusses the challenges faced by operators when selecting a policy management implementation strategy for their IMS networks and demonstrates that a unified policy management strategy is fundamental to the success for these deployments.

In order to manage and deliver premium value-added services with IMS and FMC, a policy management system is required. Policy management is fundamentally a set of rules (or policies) that describe how subscribers are supposed to receive services on the network. In IMS, two of its key components are the PDF and the Home Subscriber Server (HSS).

The HSS is the server for the network policy database, which is used to determine permissions for accessing different network services and applications, settings for how the applications and services should behave on the network, and who and when they can be used. When subscribers are authorized to use a network service such as voice or a video call, the application makes a request to the network via the PDF to reserve the network resources to deliver this service to the subscriber. The PDF is then responsible for determining, based upon the application request, what underlying network resources need to be reserved (such as network bandwidth), whether the resources are available in the network, and how to instruct the underlying network elements to reserve these resources on behalf of the subscriber using the service.

When IMS was originally envisioned, it was for a homogenous mobile network with SIP at the control layer. The PDF only needed to manage resources for a single kind of access network: GSM. Two things have been happening since the initial release of IMS — the first is FMC, and the second is that market forces are pushing service operators to offer non-SIP-based services such video on demand and online gaming.

The introduction of fixed mobile convergence creates a new set of new challenges for services providers and their policy management strategy. As operators build out their FMC networks using IMS, they will be faced with the need to ensure consistent delivery of services across these networks. The FMC network will be a heterogeneous network made up of multiple types of access networks (cable, DSL, WiMAX, 3G), with each access network using a different type of technology for transporting the service. This in turn requires mapping the policies to the underlying network. With FMC comes the introduction of nomadic users moving among the different access networks. Finally, the economics from this “network effect” introduces the need for policy peering between service providers in order to increase the scope of operators’ networks.

Market forces are pushing operators to offer an increasing menu of services on their networks. IMS was originally designed for telephony-based applications that use a call model (short duration sessions). Call model applications historically employed a centralized stateful server to manage the calls or sessions. In the IMS architecture this is the Call Session Control Function (CSCF). These call model applications lend themselves very well to using SIP for call signaling.

But, with the continued adoption of broadband, a whole new set of applications have begun to appear such as video-on-demand, peer-to-peer applications, and online gaming. These new applications do not always follow the same paradigm as call model applications and therefore do not lend themselves well to using SIP. For example, video on demand sessions have long session times and require mid-session signaling to support the video play control functions such as pause, rewind, and fast-forward. Peer-to-Peer (P2P) applications have no centralized server like the softswitch (CSCF) in the telephony applications and therefore require that the end-points to be capable of signaling for resources. While online gaming takes many forms, the multi-player online game is one of the most popular forms where many participants logged in simultaneously play against each other. These games are typically real-time in nature and many-to-many; therefore they do not lend themselves well to SIP. All of this creates the need for a PDF that is decoupled from the SIP control plane to allow non-SIP-based applications gaming to signal for premium delivery across the network alongside the SIP-based applications.

The key to meeting these challenges is a unified resource control plane in the policy management framework. Service providers, as they roll out their IMS networks, can use one of two approaches regarding the resource control plane — the siloed approach or the unified approach.

The siloed approach uses a separate PDF and resource control plane for each of the different networks. As shown in Figure 1, when the service provider migrates from a homogenous network to a heterogeneous network, it adds the corresponding PDF for the respective access network.

With this approach, the session control plane is then responsible for the handover of both the session between networks and the underlying network resources as a subscribers move across networks.

The second approach unifies the resource control plane with PDFs that signal both up and down between the applications and the network, as well as side-to-side, to seamlessly achieve resource handover between the PDFs.

With this approach, the applications only have to manage the sessions as the subscriber moves across networks and signal to the resource control plane that the subscriber has moved. The PDF is then responsible for seamlessly reserving the resources in the network to ensure that the applications get the same of level of service regardless of the underlying network. Since the PDF has a unified view of the network resources in the network it can ensure resource availability for applications.

To ensure consistent service delivery the PDF normalizes the resource control polices for each of the networks. This normalization may include the transformation of a session’s resource from a parameterized QoS model such Int-Serv to a prioritized QoS model such as Diff-Serv. And by doing so, an application does not need to be aware of the underlying network when reserving resources. It only needs to signal to the PDF the type of service it is reserving, and the PDF can determine from its policies how to reserve this for the respective network.

Finally, as operators look to deploy their policy management systems for IMS, they need to ensure that the implementation they have chosen to deploy meets scaling and performance requirements. If IMS and FMC live up to their promise, networks will become infinitely large. The unified policy management approach lends itself very well to a two-tier distributed architecture that scales both horizontally to support network growth and vertically to support network load.

A unified resource control plane for policy management ensures the consistent delivery of services across heterogeneous networks while offering a number of additional benefits to the service operators. Operators can achieve a lower overall operating cost due to fewer elements to manage and provision. At the same time, they can enjoy a faster time to market due to the simplified provisioning of new applications and the decrease in the integration efforts between applications and the resource control plane.
With a unified policy management system in place for their IMS networks, operators are positioned to begin reaping their investments. Operators can begin offering value-added services that provide a consistent high-quality user experience that stay on their IMS network.

Matt Tooley is chief technology officer at CableMatrix Technologies (news - alert), Inc. For more information, please visit the company online at www.cablematrix.com.

[Return To The Table Of Contents ]