Gijs Van Kersen suggests a way to rescue IMS from obscurity. By implementing truly network wide dynamic policy management for IMS, Web 2.0 and SOA, Van Kersen says operators can finally monetise their network
Driven by economic imperatives, mobile and fixed Service Provider networks are converging to ubiquitous IP/MPLS networks. As a result, the number and nature of applications that have to share the same bandwidth is increasing rapidly. How can operators reconcile high performance expectations with low investment budgets?
Today's IP (Internet Protocol) networks (including the Internet) are very static by nature – we provision lots of bandwidth, and then hope that it is sufficient for all traffic that wants to use it. Luckily, almost always it is sufficient. And if not, we just "drop" the excess traffic.
Which packets to drop when needed is determined by looking at a priority field in the IP packet header. Thus, only packets with a low priority get dropped (e.g. Internet data), and "delay-sensitive" traffic with a higher priority (e.g. voice, video) gets through even in case of congestion. This form of Quality of Service (QoS) is called "soft QoS" – the network will do its best to give you priority but can't make any commitments.
Internet traffic is currently growing at more than 50% per year – or doubling every two years (www.discovery.org/a /4428). A few years ago, most of that growth came from Peer-to-Peer networks – all low priority traffic that could easily cope with packet loss. However today the biggest driver is video – coming from sites such as YouTube and Facebook on the Internet; and from IPTV services on access networks.
But video is sensitive to delay and packet loss, and thus to overcome congestion will not work if 80-90% of traffic is video and voice and thus "high priority". You can compare this to buying "priority boarding" on a budget airline – what if all passengers took this option?
The PSTN (public switched telephone network) works differently. After dialling, the network will make a "hard" reservation of all needed resources along the path. Only if it can fully guarantee a connection will you hear a ringing tone, otherwise the network returns a "fast busy" signal to indicate it can't complete the call. The caller accepts this and hangs up to try again a little later.
Maybe, when a growing portion of IP traffic consists of voice and video sessions that demand good network performance, it is time to learn a lesson from the PSTN and re-introduce this hard QoS?
In fact this is what is proposed as part of the IMS (IP Multimedia Subsystem) standards defined by 3GPP and ETSI TISPAN (both telecoms standardisation bodies). Between the Applications layer and the Transport layer, they have defined a Policy & Control layer which includes a "Resource Allocation and Control Function" (RACF) to do exactly that – control and allocate resources (i.e. bandwidth and QoS) to sessions.
With IMS, when a user starts an application (e.g. a video chat with a friend), the application will request bandwidth and QoS parameters on a specific route from the RACF. The RACF calculates if this request can be properly met without impacting other applications, and if so will reserve the resources and set up the connection.
While this is a laudable initiative, it is still not sufficient. Let's see why.
Most network operators will implement both IPTV and IMS, as well as a range of other services that use the same underlying IP network. If the RACF only controls resources under the IMS domain, it only "sees" resources assigned to IMS and is unaware of resource demand on, for example, the IPTV or voice applications within the network.
For this reason, the only feasible option is the establishment of a single, network-wide policy layer, which interfaces to all applications – be it IMS, IPTV or other architectures such as SOA (Service Oriented Architecture). An end-to-end policy layer is the only method to optimise high-performance usage of the entire IP network, while ensuring that every application and every user gets the network behaviour required, appropriate to access device and location.
One of the leading service providers in APAC was among the first to realise this. They have deployed a unified network that offers fixed and mobile services including HDTV, Video on Demand and broadband internet access, all controlled on a per-session basis.
Such a unified policy layer can be an excellent source of network utilisation data, which can be mined to review network and application usage patterns for future capacity planning. And in addition to providing valuable information on aggregate habits, such information can also (with the user's consent) be passed on in real-time to specific application providers to tailor service delivery.
For example, when a user watches a video clip on his/her dual-mode phone, and enters a WiFi hotspot, the network can reserve additional bandwidth, and then signal to the application that a higher quality video stream should be sent to the user. Such enhanced user-awareness is valuable to application providers, and many will be happy to share revenues with the service provider for this.
Concluding, a network-wide policy and resource control layer will allow service providers to provide a better user experience, extract more value from the network and will enable application developers to optimise their service delivery. Marrying content and connectivity in this way will lead to a better application experience for everyone.
Gijs van Kersen Head of Mobile & Convergence Marketing EMEA, Juniper Networks