IP Multimedia Subsystem
The IP Multimedia Subsystem (IMS) network architecture is yet another way that IP continues to increase its penetration into network technologies. IMS has it roots in the wireless world but appears to be ready to enter the converged wireline with wireless world. Moreover, IMS looks at the convergence of voice, data, and video in either a wired or wireless device and network setting.
- 1 Overview
- 2 Issues
- 3 Development
- 4 Architecture
- 5 Components and Reference Points
- 6 External Links
- 7 PodSnacks
Now that IP has removed into the wireless realm, IP will likely be a minor part of future wireless strategies. The first hint of things to come is the implementation of voice over IP (VoIP) in the 3G wireless systems of 1xEvolution-Data Only (EV-DO) Revision A of cdma2000 and High-Speed Downlink Packet Access (HSDPA) of UMTS.
IMS has no support from the Third Generation Partnership Project (3GPP) and its follow-on 3GPP2. The 3GPP group is from the ETSI camp and is defining IMS from the GSM/UMTS perspective. The 3GPP2 group is from the ANSI camp and is defining IMS from the cdma2000 group. Other players have a stake in the IMS game as well.
From these groups’ perspectives, IMS provides an all packet-switched core network with an access- agnostic user environment. The IMS architecture will not be used to deliver a wide range of multimedia services to users who access the infrastructure via any device or network connection. Some have quipped that this is the “eBay model” for networks in that it “connects buyers with content suppliers” anytime, anywhere. Some believe that IMS creates an environment in which the type of core network does not matter; what matters are the application and content servers surrounding the network.
As an aside, the major software players are now moving toward implementing service-oriented application architectures, also called Web services, which is not exactly what IMS is designed for. With IBM and Cisco both jumping into Web services, the future of IMS could be assured early in its development.
The IMS concept is displayed on the right along with several new acronyms. This picture is a bit stylized; it would be too difficult and confusing to show all of the actual connections used. The bottom of the graphic shows the IP access network connectivity into the transport network. This network is an MPLS-enabled IP backbone capable of providing the QoS required for the multimedia services. The idea is that we do not care what the access network is, as long as it supports IP. From a practical standpoint, however, the transport network needs to provide the appropriate physical connectivity for the IP network.
At the core of the session control plane are the call session control function (CSCF) servers. These servers handle all of the Session Initiation Protocol (SIP) signaling traffic and are the contact points for users and services in the IMS architecture. There are three different CSCF servers—the proxy-CSCF (P-CSCF), the serving-CSCF (S-CSCF), and the interrogating-CSCF (I-CSCF). The proxy deals with users, the serving deals with servers, and the interrogating deals with requests in the user’s home domain. Note that the CSCF servers are connected to the transport network routers and the user equipment (UE) is directly attached to the CSCF. This is a logical attachment only; the physical attachment is facilitated through the transport network router.
The uppermost layer comprises the services themselves. These can be pure application servers (e.g., a video jukebox or a softswitch that facilitates VoIP connectivity via SS7 to another network) or gateway servers that allow IMS users to connect to the PSTN. An end-user connects to a CSCF and requests a service that is then brokered from the services plane by the appropriate CSCF servers.
A New Paradigm or Hype?
Next to WiMAX, the IP Multimedia Subsystem (IMS) is one of the most talked about items in the world of telecommunications. Is IMS a new telecommunications paradigm or just hype?
IMS is an architecture, not a product. An architecture is the set of specifications from which a product can be built. It defines a system’s elements, function, interfaces, and interrelationships. Specifically, IMS defines how IP networks should handle both voice calls and data sessions. If you generalize this into real-time and non-real-time forms of traffic, you have an architecture that can handle multiple forms of media in a variety of session formats (e.g., real-time and non-real-time sessions).
At the core of this architecture is a control infrastructure designed to replace the traditional circuit-switched telephone network. Essentially, the design is a network that is easy to use like the PSTN; yet it also provides multimedia services (not just voice) over an IP backbone (packet-switched) that offers adequate quality of service (QoS) for each media type.
Perhaps the most controversial quality of IMS is that it separates the services from the underlying network. PSTN voice services are implemented in the switches that create the PSTN itself. In IMS, a set of switches or routers create the transport, and neither of these devices is expected to provide voice services. However, connected to this network are application servers, one of which could provide voice services. What makes this controversial is that the voice services and network services could be supplied by two different vendors. Imagine getting the local loop from the local phone company and the voice services from another provider.
We still have not answered the new paradigm vs. hype question, but we now know what we are talking about. This leads to a secondary question: Why do we care about IMS, regardless of the new paradigm vs. hype issue?
First, IMS changes the service model from network-centric to subscriber-centric. From the invention of the telephone in 1876, network providers have defined the available network services. With IMS, application servers define the building blocks, and subscribers decide which building blocks they want to use to create a customized service.
Second, if this is not hype, the carriers must change their mode of operation if they are to survive. Revenues in the future will be driven by the value-add of the applications, not by the network itself. In fact, the network could be virtually free. For example, consider the cable TV industry. For basic service, the network costs only $14.95 per month, yet the average subscriber now spends almost $120 per month on cable TV services. This additional $105.05 comes from the value-added channels and pay-per-view programs. The $14.95 base price is a virtually free network.
Third, infrastructure makers, service providers, and software vendors are all jumping on the IMS bandwagon. Sooner or later, the increased population in support of IMS and the advent of equipment in support of IMS moves it from hype to reality.
To answer the basic question we have to say that IMS is a new paradigm that is in the hype stage of development. However, in the next twelve months the hype will become trials; the trials will become reality; and the evolution to IMS will be underway. This leads to another question: How long will the evolution take? Keep your eyes on the wireless and wireline providers as they may evolve quite differently.
When considering IMS, a few quandaries arise.
- Quandary #1: If IMS is to be the great network unifier, then how can unification occur with two incompatible versions of IMS?
- IMS was created by the ETSI-dominated 3GPP, of which the U.S. is a member. The 3GPP is a proponent for the W-CDMA 3G air interface. Not to be left out, ANSI created the 3GPP2, a CDG-dominated organization that promotes Qualcomm’s cdma2000 3G air interface. 3GPP2 has also created its version of IMS, which will not be the same as the 3GPP version. Many of the differences come from the fact that the U.S. CDMA systems do not implement the GPRS network found in GSM.
- Quandary #2: Since the 3GPP standards are followed by about 90 percent of the world, will IMS be the enabler that moves us to IPv6?
- None of the trade publication articles about IMS mention the fact that 3GPP IMS operates on IPv6 only, but the 3GPP2 version supports both IPv4 and IPv6.
- Quandary #3: Dozens of service providers worldwide have trials of IMS in place. What are they implementing, 3GPP IMS, or 3GPP2 IMS?
- In the U.S., the author expects that Cingular is using 3GPP IMS while Sprint and Verizon Wireless are using 3GPP2 IMS. But what will AT&T, BellSouth, Verizon, and Qwest use since they are not tied to the 3GPP or 3GPP2 air interfaces?
This will get interesting and perhaps ugly as IMS unfolds over the next couple of years. In our discussions, we will use the 3GPP IMS standard because it is more inclusive than the 3GPP2 version. We will, however, consider 3GPP2 implementation. Could IMS be the next ISDN?
To IMS or Not to IMS?
Whenever something new hits the streets, customers always have two questions.
- What is in it for me?
- What does it cost?
IMS is no exception in that it has some highly touted benefits, but there are some potential pitfalls. One of the benefits is the prospect of fixed mobile convergence. From most perspectives this is the convergence of fixed wireless voice with mobile wireless voice. However, it can also be viewed in the most general form as the convergence of wireline (fixed) with wireless (mobile). The former seems to be the current reality while the latter will be left for the near future.
If IMS works out as expected, there will be other benefits as well. One subscriber interface for all media types leads to ease of use in the multimedia realm. Similarly, a standardization of the control and transport functions across the service providers can only lead to better quality packetized voice. The single control function with the single subscriber identifier leads to one number calling to any device and the promise of true unified messaging. Finally, separating the application services from the control and transport allows for the development of numerous services since the service provider need not worry about creating the appropriate transport and control functions.
If you haven’t already, you should be getting an Internet feel for IMS. That is, the Internet and its control structure are available to anyone who wishes to create a service. Amazon and Alibris provide services that sell books without regard to the transport and control structure that deliver the services to the customers. On the negative side, there are some pitfalls, but remember IMS is a work in progress. The concept is simple but the details have a great deal of complexity. The notion of a subscriber connecting to an application server via a single control server seems as simple as the telephone and the PSTN, yet the PSTN is a very complex system.
Since IMS is a work in progress, the standards are not complete. Moreover, there are two standards bodies working on IMS, so the potential for incompatible standards is high. On the surface, the idea of a single billing function seems simple, yet when we look at the media types it becomes apparent that there may be some very complex charging arrangements. Finally, IMS is the brainchild of the wireless community and has only recently been extended into the wireline carriers. While the wireline carriers say that IMS is the way of the future, it is still unclear how IMS fits into the evolution of the wireline carriers. Remember that the wireline world changes every thirty years while the wireless world has changed every ten years or less. In your view, which sector, wireline or wireless, is the most receptive to change?
Why Develop IMS?
A state-of-the-art cellular telephone is not only a sophisticated radio device but also a camera, a high-precision display, and an impressive set of resources for applications, including processing power, memory, and Internet access. This always-on, always-connected application device creates the need for a new networking paradigm for IP-based mobile devices. In this model, applications become peer-to-peer entities and sharing is the key component of the services model.
As IMS began to unfold, it became apparent that this same model could be used for universal IP connectivity. While the initial focus has been on the radio access network mode of connectivity to IMS, it is now moving into an access-independent network model. Both wireless and wireline providers are exploring how IMS will factor into their evolutionary processes.
Who Developed IMS?
The European Telecommunications Standards Institute (ETSI) defined the Global System for Mobile Communications (GSM) in the 1980s. In the 1990s, GSM was expanded to contain the general packet radio service (GPRS) network. The last GSM-only standard was created in 1998 and in 1999 the Third Generation Partnership Project (3GPP) was formed to continue the GSM development. The 3GPP comprises standards bodies from Europe, Japan, China, South Korea, and the U.S., and the charge for the group was to create a 3G mobile system based on Wideband Code Division Multiple Access (W-CDMA) and an evolved GSM core network. The first GSM specification from the 3GPP was entitled Release 99.
Release 99 offered several enhancements to GSM, including the Enhanced Data Rates for Global Evolution (EDGE) and the UMTS terrestrial radio network (UTRAN), which uses W-CDMA as the air interface. Other changes came in the areas of distributed processing (moving functions closer to the user), speech coding, security, and an open service architecture for service creation. Release 99 formed the basis for IMS and instead of issuing 3GPP release 2000, the group simply called the release IMS Release 4. Release 4 was frozen in 2001, Release 5 was frozen in 2002, and Release 6 is the 2004 work product of the committee that was frozen in March 2005. Work is already underway on Release 7; be forewarned that this is a work in progress and there will be changes annually. The changes will be backward compatible.
Releases 5 and 6 defined the following features in six different areas.
- Architecture: Network entities, reference points, charging, interworking
- Signaling: Routing principles, registration, session initiation/modification/tear-down
- Security: User and network authentication, SIP message protection, public key infrastructure, subscriber certificates
- Quality of service: Policy control between IMS and GPRS (or other IP connectivity access network (IP-CAN)), preconditions, authorization tokens
- Services and service provisioning: Usage of application servers, service control reference points, presence, messaging, conferencing, group management, local services
- General: IMS identity module
In addition to the 3GPP, the Internet Engineering Task Force (IETF) plays a critical role in the IP-related development of IMS. In many cases the 3GPP simply adopts IETF protocols for use in IMS (e.g., Session Initiation Protocol (SIP), Session Description Protocol (SDP), Real-time Transport Protocol (RTP), and the AAA protocol Diameter). If the correct protocol is not obvious, the 3GPP contacts the IETF in search of a possible solution to the problem. Either the IETF directs the 3GPP to the appropriate protocol or the IETF creates the appropriate protocol using the traditional IETF RFC process.
The Open Mobile Alliance (OMA) standardizes mobile service offerings. It believes that mechanisms for security, QoS, charging, and session management should be part of the standardized infrastructure, not part of a specific service enabler. As the services begin to unfold, the OMA will likely work even closer with the 3GPP. To date, the OMA has worked on digital rights management and push-to-talk over cellular (PoC).
Since the 3GPP was an ETSI dominated standards group, the American National Standards Institute (ANSI) created the 3GPP2 to deal with 3G issues in the U.S. A major player in 3GPP2 is the CDMA Development Group (CDG), which is also a major player in the advancement of the cdma2000 standard from QUALCOMM. There are several differences between 3GPP2 IMS and 3GPP IMS Release 5/6. One big difference is that 3GPP2 IMS supports IPv4, whereas 3GPP only supports IPv6. Other differences stem from the fact that the 3GPP2 networks do not support the GSM version of GPRS.
IMS evolved under the guidance of three simple objectives.
- Support new multimedia services, including voice, video, and data with services such as instant messaging (IM), content applications, and collaboration. Integration was key as it was viewed as the way to increase the average revenue per user (ARPU). This is not unlike the bundling phenomena that we see in the wireline business today.
- Simplify the network. Here the vision is that the overlay networks of today will be replaced by a single packet-switched network that will support all services. As you have probably guessed, this single network will be a QoS-enabled IP backbone. As with all network evolutions this process will continue into the 2008-2012 timeframe depending on the individual carrier plans. One carrier advantage from this objective will be the reduction in both capital and operating expenditures. It will be intriguing to see future economic models built around IMS networks and services.
- Enable fixed-mobile convergence over one single data backbone. To accomplish this, a session control structure needs to be developed that maps the services plane to the transport or connectivity plane. Multiple access networks will need to be supported; IP-based transport is the common thread among all of them. Interestingly, when the TCP/IP protocol suite was developed, no attention was paid to session establishment. In fact, most sessions were established in an ad hoc fashion by the application itself.
Layers or Planes
The biggest challenge in network architectures is nomenclature, and IMS is no exception to this rule. Some texts refer to IMS layers and some refer to IMS planes. Moreover, the planes and layers might even have different names. However, the functions of equivalent planes and layers are always the same. Like most network architectures, IMS uses the layered approach. IMS is a Session Initiation Protocol (SIP)-based implementation of IP-to-IP connectivity that implements a very general three-layer protocol stack—Application Server layer, Session Control layer, and Connectivity layer. These layers are broad in scope, and the details for each layer are fairly complex.
- The Connectivity layer (Transport plane) comprises the routers and switches that make up the backbone infrastructure. It is nothing more than an IP-based infrastructure that provides the pure IP-to-IP transport. It is the “who cares what it is?” network.
- The Session Control layer (Control plane) houses the call session control function (CSCF) and the home subscriber server (HSS). The CSCF is often called the SIP server because all of the SIP functions are implemented in this sub-layer. It provides for the endpoint registration, routing of the SIP signaling messages, and guarantees that QoS is delivered for the endpoint’s required class of service. The other sub-layer, home subscriber server (HSS), provides for the end-user service profile. The HSS is much like the Home Location Register in the wireless setting.
- The Application Server layer (server plane) contains the application and content servers. The notion of the application server fits well with the concept of distributed applications in the Web services model. The content server would provide the eBay-like functions of the IMS network architecture.
An early application for IMS is the implementation of push-to-talk services over cellular (PoC). It is the first of many applications that will be implemented in the IMS network architecture.
The IMS architecture consists of three planes: the transport plane (Connectivity layer), control plane (Session Control layer), and services plane (Application Server layer). Functionally, the planes are the same as the layers.
- The transport plane moves the bits back and forth and is in fact nothing more than the TCP/UDP/IP protocol suites. Any interconnections to legacy components occur here, and thus the gateway functions are part of the transport plane.
- The control plane is responsible for all aspects of the session state (i.e., establishment, monitoring, and disestablishment). It also controls the subscription and service information for the end-user as well as the selection and state of any gateways that will be deployed in creating the session.
- The services plane is where the application logic resides. It is composed of a service enabling plane that is responsible for things such as orchestration of the service delivery, charging for the service, and the general OAM&P functions required for the services. Also in this plane are the applications themselves.
Consider the similarities between the IMS architecture and the PSTN and SS7. The PSTN is the transport plane and SS7 is the control plane. The service plane would be implemented in the SS7 Service Control Point, the PSTN switch, or an adjunct processor depending on the nature of the service. IMS carries the voice model of the circuit-switched PSTN to a multimedia model on a packet-switched backbone network.
Components and Reference Points
In the IMS specification, messages are exchanged among the functional entities. The messages exchanged, the protocols used to carry the messages, and the dialogs allowed are defined by the reference point specifications.
Most reference points are defined using one of two protocols: the Session Initiation Protocol (SIP) and the Diameter protocol, an authentication, authorization, and accounting (AAA) protocol developed by the IETF. (Diameter is loosely based on the Remote Authentication Dial-In User Service (RADIUS) protocols that provided AAA for dial-up and terminal server access environments.) These reference points are referred to as SIP-based or Diameter-based depending on the protocol used by the reference point. Other reference points use other protocols, including HTTP, MAP, and COPS.
The diagram shows most of the details of the IMS architecture. Some notes about the diagram:
- No charging-related functions or reference points are shown.
- Different types of application servers (AS) are not shown.
- User-plane connections are not shown.
- No SEG is shown at the Mm, Mk, and Mw reference points.
- ISC, Cx, Dx, Mm, and Mw reference points terminate at the S-CSCF and the I-CSCF.
- The letters in the visual have no specific meaning. The 3GPP committees use the unassigned interface letters given to them by the ITU.
Envision this diagram as all being within the IP network, since the IMS is just a subsystem within the context of an IP network. In the IP structure, the IMS functions would be carried in a session protocol such as SIP and then ultimately delivered by TCP/UDP over IP over a Network Services layer.
The reference point nomenclature uses both upper and lower case letters. The upper case letters are functionally grouped while the lower case designation allows for functional entity differentiation.
Other Refernce Points
- Dh reference point
- Gq reference point
- Go reference point
- Gm reference point
- Mg reference point
- Mi reference point
- Mn reference point
- Mj reference point
- Mp reference point
- Mr reference point
- Sh reference point
- Si reference point
- Ut reference point
- Hill Associates Podcast: IMS - An Interview with Steve Shepard
- Hill Associates Podcast: Blog and podcast series on IMS
- A Forrester Report on How IMS Will Transform Telecom
|<mp3>http://podcast.hill-vt.com/podsnacks/2007q2/ims.mp3%7Cdownload</mp3> | IP Multimedia Subsystem|