In one regard, Quality of Service mandates for voice and multimedia over IP services are nothing new: failure to meet them can quickly result in lost customers and revenues. Delivering profitable multimedia services means giving customers exactly what they expect and pay for every time they use the service.
In another regard, sustaining high-quality services represents a whole new arena for service providers where they are confounded by issues unique to IP infrastructures every day as subscriber numbers grow. Fraud, evolving regulations, and the idiosyncrasies of IP itself all pose challenges that must be met in order for services to keep pace with the performance of the PSTN and improve upon its profitability.
Outwardly, Quality of Services simply means delivering the consistently high levels of performance end users routinely expect. Internally, sustaining QoS is not so simple. Service providers must manage many different facets of the communications path as a call progresses from the end user through the service providers network and potentially into a third-party network. In IP-based multimedia services, quality must be achieved and ensured using broadband networks that were never intended to handle real-time traffic with such predictable precision.
A closer look at todays networks reveals numerous factors influencing the successful delivery of profitable high-quality services. To start with, two unique facets of IP multimedia infrastructures must be countered: the bursty nature of data networks and the elastic nature of IP connections.
Broadband Access: Burst Versus Best
The market for broadband voice and multimedia services is growing rapidly. iLocus estimates that by mid-2005 there were 14.5 million consumer voice over broadband subscribers, and that the number of subscribers in the top 10 countries alone will exceed 130 million by 2009.
But despite an apparent abundance of bandwidth, most access networks were designed to deliver the required service levels to consumers based on the bursty nature of data services. Networks could be overbooked and still deliver the required service so that, in the case of downloading an MP3 or video clip, for example, a one- or two-second gap in the stream would go unnoticed. (However, the same one-second gap in a voice stream would be fairly annoying.)
In other words, although a connection rated at 1Mbps may seem to be more than sufficient in making a voice call, the service provider could face the thorny problem of all or most subscribers trying to make multimedia calls over the access network at the same time. Without adequate planning and precautions to ensure QoS, such an occurrence could compromise quality, not to mention SLA compliance and profits, fatally.
IP is Elastic (Voice is Not)
The second factor in the QoS equation is the elastic nature of IP connections. In the traditional telephony world, if you have a connection capable of supporting 32 concurrent calls, it does exactly that the thirty-third call will not fit and the caller will get a busy tone. In the world of IP, because each connection is just a stream of packets, the thirty-third call will be admitted too to the detriment of all 33 calls. Each IP stream will suffer from some delay and dropped packets resulting in a reduction of voice quality. Blocking a call may seem harsh but the alternative is 33 unhappy customers!
Enter QoS: Imposing Order on Chaos
Combining the inherent overbooking in the access network with the tendency of IP streams to compete for available bandwidth sounds like a recipe for disaster. In response, service providers and infrastructure suppliers continue to evolve comprehensive strategies and innovative mechanisms for imposing order on would-be chaos.
Policing the Access Network
The access network will not police itself, so it is necessary to impose some control on who can do what and when. Doing so requires an understanding of the relationships between the different parts of a multimedia call. Each call is made up of a signaling stream and a media stream. When most network devices encounter these streams they are blissfully unaware that there is a relationship between them. This is where Session Border Controllers (SBCs) come in. SBCs, as their name implies, are session aware, meaning they understand which media stream belongs to which signaling stream.
This in turn means that, unlike most network devices, SBCs exert control on the traffic at the level of an individual session when required. First developed to provide secure NAT traversal, SBCs have evolved to encompass such a broad range of security and quality management functions that they are rapidly becoming indispensable in creating secure, manageable multimedia networks.
For starters, SBCs now furnish the nuts and bolts of quality management in the access network:
Session Admission Control: understanding the topology of the access network and the traffic flowing through it.
Controlling parasitic traffic: preventing unauthorized use of signaling and media streams.
Resisting Denial of Service attacks: blocking or limiting malicious IP or signaling traffic.
QoS policing and remapping: ensuring correct priorities are assigned to traffic
Anti-tromboning: preventing unnecessary media traffic in the access network.
Session Admission Control
The basic principle of session admission control is preventing the admission of more calls than the network can handle at any given time. At the simplest level, Session Admission Control involves modeling the bandwidth in the access network and monitoring the resources consumed by each new call. Once the limit is reached, further calls can be refused, thus guaranteeing the quality of existing calls.
At a more detailed level, a corporate customer with an IP PBX may be permitted to make up to ten concurrent calls equivalent to ten external lines on the PBX. Once the agreed limit is reached, further calls can be refused. However, one important exception must be made to the policing: If an emergency call is received it will always be carried, regardless of any restrictions.
As a simple example of service theft, a user might signal that a voice call is being made, then initiate the exchange of high-capacity video data once the path is established. This hits the service provider on two fronts: a) loss of revenue by billing for only a voice call, and b) potential degradation in service quality for other users resulting in dissatisfaction.
The structure of a VoIP call with separate media and signaling streams can lead to some innovative ploys. For example, rogue PC clients can transport media in the RTCP quality monitoring stream without ever being policed in most networks. Another ploy involves transporting media in the call signaling stream then failing the call before billing commences resulting not only in unbilled calls but in repeated call sets that can cause huge signaling rates that themselves constitute a DoS attack.
To counter creative new strains of fraud, SBCs and other devices must police all components of a call to ensure that the call is executed as requested and RTCP traffic is within expected bounds.
Resisting Denial of Service Attacks
All devices that support voice and multimedia over IP are subject to DoS attacks including traditional IP logic and flood attacks as well as newer exploits at the SIP signaling level and application level. While conventional firewalls go part way toward preventing IP level attacks, firewalls are still typically session-unaware and unable to recognize SIP level attacks.
Here it falls to devices such as SBCs to prevent the propagation of disruption in access networks. Measures must be taken to keep malicious traffic out of the network, thus preventing it from becoming clogged. Service provider infrastructures must be shielded from being overwhelmed by limiting signaling rates to levels that Softswitches in the core can handle.
QoS Policing and Remapping
All calls carry information in IP packets that determines the priority a packet receives as it traverses the network. Service providers must ensure quality settings are enforced so that they can assign the correct service to different types of traffic and prevent some users from manipulating the settings at the expense of other users. True carrier class SBCs equip providers to control quality settings based on traffic source, type, and even individual users.
This function is invaluable for interconnection of service providers networks as well since QoS policies and policing capabilities may not match the settings used by a peering providers. Here, the SBC steps in to remap quality settings as traffic is exchanged.
This strange term refers to the trombone-like shape that a call path takes when a local call is made via a distant SBC. When both end points of a call are in the same network, it may not be necessary or desirable to route the media via the SBC. When the SBC detects that the call can be more efficiently routed directly between the end points it will release the media, thus freeing up valuable bandwidth in the access network.
Regulation of Services
We have already mentioned that networks must provide Emergency call capabilities under all conditions, and evolving regulation of VoIP and MoIP services increasingly requires operators to provide Lawful Intercept capabilities as well. While this may not be seen as part of the quality management of the network, the imposition of these requirements means that inappropriate or illegal activities are more likely to be discouraged. Implementing lawful intercept capabilities demands that service providers know about all the signaling and media flows in their network. Once again, the SBC is a natural candidate to support these requirements.
Premium on Performance = Premium on Profit
Consumers of voice services are fickle and dont like to encounter the unexpected. Carriers are literally betting the business on their ability to strike the right balance between cost, quality, and control in provisioning access networks. Over-provisioning the access network doesnt translate into a reliable service delivery platform in the face of bursty data traffic.
Only a multi-faceted approach to policing available resources can ensure customer confidence and lock in loyalty by delivering consistent levels of service, day in, day out. Providers must anticipate and engineer for varying traffic patterns, employ contingency planning that sustains customer communications while upholding the law during emergencies, and literally put their money where their mouths are, in terms of SLAs and fraud prevention.
As global communications migrate toward all-IP infrastructures a little each day, service providers continue to seek ways of rapidly embracing a more profitable future while preserving the legacy of quality taken for granted in the past and present. Fittingly, as the realities of VoIP become clearer and more present each day, a balance between innovation and tradition is being struck in a level of the network for which the PSTN has no counterpart.
And if service providers and their partners do it right, end users need never even know this level is there, even as they gladly pay a premium for the quality it guarantees. IT
Mike Wilkinson is vice president of Marketing for Newport Networks. For more information, please visit the company online at www.newport-networks.com.
If you are interested in purchasing reprints of this article (in either print or PDF format), please visit Reprint Management Services online at www.reprintbuyer.com or contact a representative via e-mail at [email protected] or by phone at 800-290-5460.