Research analysts, vendors and the media have been touting the growth of the voice over IP (VoIP) market and the value inherent in voice applications for some time. In fact, the Yankee Group earlier this year stated that they see the business VoIP market reaching $3.3 billion in service revenue by 2010.
There is a lot of hype around this technology, and deservedly so, but from a user perspective, VoIP is getting a less-than-stellar reputation. If you look beyond high-profile implementations and examine most IT organizations, there is a hesitance to move from the project stage to deploying VoIP enterprisewide because there is so much at stake if something goes wrong—many businesses can’t afford any network downtime due to bandwidth drain.
Is this a case of industry watchers and vendors over-inflating facts and stretching the truth? Not really. The fact is that while VoIP is a technology whose time really has come, there is a perception in the market that it simply does not work. Organizations relying on traditional data or network management solutions to manage their voice applications, rather than employing a tool specifically designed for that purpose, will be less than satisfied with the performance of their VoIP deployment. This problem creates a lose/lose situation for both vendors and enterprise IT departments, in which the technology does not deliver the anticipated results and IT decision-makers lose credibility within their organizations.
Businesses need to think out of the box when it comes to managing VoIP on their network. The reality is that old solutions can’t solve new problems because the old set of best practices is no match for the new challenges associated with VoIP. To properly manage VoIP and the new challenges presented, organizations need new processes and the development of new skills.
Going Beyond Fault Tolerance
Most IT departments manage their network reactively based on outages or when its fault management platform discovers poorly performing equipment in the network. This black-and-white method of managing does not account for a data and voice convergence paradigm, with shared infrastructure supporting multiple types of traffic. Monitoring connectivity between two end points is not enough to qualify the quality needed to successfully support VoIP. In other words, IP is about shades of gray, and you need to have management tools that will poll the network infrastructure in an ongoing manner. Tools that don’t deal with shades of gray are too reactive and only focus on the performance levels of the network infrastructure, not proactively foreseeing potential issues.
Managing VoIP with traditional network solutions won’t work, as these resource-centric tools do not grasp the concept of the end-user experience. Reports from these solutions are geared for a technical audience, with the bottom-line goal being that the network is up and running. This kind of service monitoring only reveals packets of data lost and does not paint a picture about customer or user satisfaction. At the end of day, IT organizations need to turn metrics into an assessment of service quality. How do you identify customer expectation? How do you compare how performance is being delivered on an expectation level versus customer experience? You certainly can’t capture it in traditional fault tolerance solutions—on a service level, you won’t get the big picture on user experience, only reports on performance.
For example, in a VoIP environment, if there is a one-hour conversation with one minute of poor voice quality, network managers relying on fault management solutions may not even sense a problem since no downtime was detected and any data loss fell into an acceptable range of fault tolerance. However, anyone who has experienced a dropped call or inconsistent service knows that even one minute of bad call quality is unacceptable and does not happen in the legacy phone world.
VoIP is a real-time application that needs to be treated as such—it does not have the same kinds of requirements as other enterprise applications that IT departments are used to managing. To meet end-user quality of service expectations, VoIP traffic must be prioritized over less mission-critical data traffic using router-based quality of service (QoS) and class of service settings. Poor capacity planning and bandwidth allocation between data and voice applications can lead to “quality of call” issues if all the available bandwidth is dominated by data applications or, at the other extreme, if VoIP traffic overwhelms the bandwidth need for business-critical applications. As most networks are in a constant state of change, IT departments must be able to monitor and manage the QoS and class of service performance for both the VoIP-specific virtual LANs and the data virtual LANs on a day-to-day basis.
QoS management has been done for years, but under a best effort paradigm where it was essentially a prioritization mechanism between low priority traffic and business critical applications. QoS management in the context of VoIP takes that paradigm to the next level—the desired end result is moving from best effort performance to assured performance by class of service. A best effort strategy is not enough to guarantee quality of experience.
Navigating Potential Pitfalls
IP itself is a finicky protocol and requires a very particular approach. Although most of its eccentricities are manageable for data traffic, it wasn’t designed to support real-time applications like VoIP, and requires careful monitoring and management to meet business end-user expectations. The three metrics that can negatively affect VoIP are jitter, latency and packet loss. The degradation of any one of these can turn a good call into unintelligible garble. All three metrics are well known to IT departments, but VoIP adds a new layer of requirements to them, including the need for real-time resolution and end-to-end monitoring.
Network administrators need this level of granularity to detect problems that are transient in nature and may only happen one minute every hour. These metrics need to be tracked in a manner that accurately reflects the end-user experience. When managing a large-scale network with hundreds of critical sites and thousands of minor ones, validating VoIP performance means validating how each site works together. Administrators need to know how each site individually works with all the others and this process needs to be an ongoing assessment of communication between all sites to ensure a high quality experience.
The way for enterprise IT to track jitter, packet loss and latency is by using a fully meshed strategy, including intra-site and site-to-site, which enables them to verify that the enterprise network is always capable of transporting voice traffic. In addition, the IT department needs a mechanism in place to verify and detect any blips in VoIP network readiness, even when the majority of the end-users are offline. This can be done by policing and shaping network traffic.
Policing and shaping network traffic is critical, and organizations need to leverage advanced QoS techniques and features to do so in order to prevent congestion. This must be augmented by a QoS policy that will limit maximum traffic for other applications. This kind of management cannot only be done at link level, but must be approached in a systemic manner throughout the entire network to guarantee overall consistent quality of experience (QoE) over the network infrastructure. IT managers don’t want to create great call quality in some instances with bad quality on other occasions. Consistent quality of experience must be defined with a systematic approach based on a distributed management system.
Achieving Quality User Experience
The new mandate of network managers is service delivery. To achieve a quality user experience with VoIP, reaching new levels of performance monitoring and performance assurance is a necessary step in the evolution of performance management. While monitoring and keeping track of all network activity have been traditional goals for network managers, today’s VoIP requirements call for an evolved performance management discipline that will manage quality of service and quality of experience in different way.
There are many variables to consider in this new management paradigm: What is the service throughout the network? What are VoIP and packet delays between different combinations of endpoints? Where does trouble shooting fit in ? If there is a problem between two end points, you need to know where, and the location of the weakest link in delivery chain. Just knowing how many packets are being dropped is not enough. The only way to achieve this is not with probes, but with end-to-end measurement and QoS management. Even QoS by itself is not enough. The quality of a user’s experience must be managed at the application level. In terms of granularity, the performance between two locations and quality of experience between users need to be validated. Having greater granularity at the application level provides more insight into the user experience and quality.
The Future of Performance Management Is Service-Centric
Using a service-centric performance management approach such as this is a necessary starting point in a converged network, as it brings together multiple pieces of the puzzle and helps IT understand the roles played by the different teams around service delivery. These service-centric solutions turn technical metrics (jitter, packet loss) that don’t tell you much about assessment of call quality into metrics that can be used to ensure the end-user experience is as close to flawless as possible. Next generation performance management solutions apply state-of-the-art algorithms to determine mean opinion score that tell you if call quality is good or not.
Voice over IP is a technology that should not be managed by point solutions with regard to user experience. To be able to assess VoIP performance, you need to not only monitor and manage the performance level of each component, but need to validate how they interact together. That level of interaction is key. Network managers must keep an open mind to find new solutions to new problems, while working with the current paradigm shift to accommodate for the future of converged networks.
Deepak Bhargava is product manager, enterprise market, and Bruno Zerbib is director of product management, network solutions for InfoVista.
For related stories, see: