A VoIP management system should help to ensure that a high level of call quality is maintained. One survey of 250 IT executives by Network World found that VoIP quality-of-service assurances were the number one VoIP drawback. In an odd twist, call quality is a source of enormous stress in a VoIP deployment because it is a nonissue in the PSTNit is something you just don't worry about. Before beginning a VoIP deployment, many consider it an unknown factor.
Two broad areas affect call quality:
Problems at the VoIP server Overloading, constraint of a key resource, or failure of a critical component
Problems in the network Configuration errors, congestion, attacks
By using the planning guidelines discussed in Chapter 3, you should not experience poor call quality on an initial deployment. However, things will no doubt change as new users and applications are added to the network. Good management practices can ensure that call quality is not an issue as the network traffic and number of users increase. Monitoring VoIP components is a way to help ensure that the experience for end users remains good.
Determining the Call Quality
The first step in managing call quality is to determine what level of quality your users are experiencing. Call quality is measured in terms of a MOS. MOS was discussed in great detail in Chapter 3, but as a brief refresher, remember that the scale ranges from 1.0 to 5.0, where 1.0 is very poor and 5.0 is excellent. A MOS of 4.0 or better is considered toll quality. A MOS below 3.6 could be considered poor and would not be good enough for most business-quality phone calls.
How do you know what the MOS would be for calls on your network? Management tools can monitor network performance statistics and calculate a MOS based on these statistics. Following are some guidelines for monitoring call quality:
Monitor continuously It is important to monitor 24 hours a day, 7 days a week. By doing so, you can establish network baselines for quality and spot trends in the call quality. Trends can help you determine if the quality is declining over time and, if so, the severity of the decline.
Monitor different areas of the network Monitor network links where the VoIP traffic flows. WAN links for branch offices are especially important areas for monitoring call quality because of their limited bandwidth and variable usage patterns.
Monitor actively If you have management agents that can actively simulate calls to monitor their quality, you can monitor even while no users are making calls on the network. Active monitoring lets you be proactive and spot potential call quality problems before your users experience them.
Monitor passively Most IP PBX systems can provide some quality statistics for the calls they control. These statistics are usually available in CDRs, which report measurements from already-completed calls. With passive monitoring, you know the quality after the fact, which does not help prevent quality problems, but may help with problem diagnosis and resolution.
Management tools generally enable you to set call-quality thresholds. When the call quality drops below the threshold, an action can be taken to alert the appropriate staff or perform other automated tasks. And these tools often offer reporting capabilities. Reports that reveal call-quality metrics over time are useful when looking for trends. Call-quality trends can help you to spot problems before they affect end users. Figure 6-4 shows call-quality trends over time.
Figure 6-4. MOS Changes Among a Set of Locations, Across Four Days
Chapter 3 showed that call quality equals network performance. If the network performance is bad, then the call quality will be poor. A good VoIP management strategy includes network performance management.
Managing Network Performance
Network performance issues are quite often the culprit behind call-quality and availability issues. By defining and then monitoring performance metrics for the network, network performance management helps to avoid problems, or at least reduce their duration. Monitoring needs to occur on a continual basis to ensure good network performance.
Performance management not only ensures good call quality, but also measures how VoIP traffic affects the other applications running on the network. Most business applications that use TCP have performance requirements much different from VoIP. For example, TCP applications usually require low network response time and high throughput. Does the response time increase dramatically as VoIP traffic is added to the network? Does the throughput drop when the application begins competing with VoIP calls? These are the kinds of questions that network performance management can help answer.
For VoIP, the key network performance metrics are delay, jitter, and lost packets. The MOS is directly affected when any of these statistics increases. Although many network components may be active between two IP phones, the overall network performance is only as good as the weakest links. Network performance management therefore should look at the key performance metrics for the entire path to see the level of quality that your users are experiencing.
A network performance management plan includes service-level management and QoS management. The following sections examine what is involved with each.
Users need to be as happy with the level of service being delivered as the IT team. Service level agreements (SLAs) provide a quality target for the actual performance you are delivering.
To create a VoIP SLA, you define a VoIP performance target, using the best information you have available. Start by taking baseline measurements, to make sure that the performance levels you seek are actually achievable. Historical performance data also can serve as a baseline of normal operating characteristics and use of network elements and end systems. For example, you might define an SLA that says the MOS should be above 3.9 for 98 percent of the core business hours.
Finally, collect the key performance metrics and compare them with the SLA. Performance data gathered on an ongoing basis provides good input to help effectively plan for infrastructure growth. And because an SLA is often put in place to manage a service provider, you need to define what levels of service are acceptable, what happens when the agreement is not met, and how the service will be managed.
SLAs for VoIP are usually structured around the key VoIP performance metrics: availability and MOS. The MOS is determined by the following (and the codec being used):
Delay VoIP traffic is intolerant of delay. Long delays can make phone calls sound like walkie-talkie conversations. SLAs for delay are usually specified as a maximum allowable time, in milliseconds, for packet delivery.
Jitter Variations in packet arrival time can cause packets to be discarded and VoIP call quality to suffer. SLAs for jitter are usually specified in maximum allowable milliseconds of variability in delay, among packets transmitted from the same source.
Lost packets If lost, VoIP packets are not retransmitted. Lost packets thus result in missing syllables or words in a call. SLAs for lost packets are usually specified as a maximum allowable percentage of all packets sent.
Chapter 7, "Establishing VoIP SLAs," discusses SLAs in greater depth.
Another aspect of managing call quality involves the management of QoS. The goal of QoS management is to make sure that QoS is configured correctly and continues to work correctly.
Chapter 5, "Quality of Service and Tuning," discussed the complexities of QoS configuration. Implementing and configuring QoS to do exactly what you need for your network can be a daunting task. There are the many different QoS mechanisms to understand and, unfortunately, not all vendors implement them exactly the same way. In addition, instead of talking in terms of bit fields and traffic classes, you probably want to define your QoS configuration in terms that are familiar to your enterprisesuch as gold, silver, and bronze levels of service. Policy-based network management, touched on briefly in Chapter 5, can provide help with network QoS configuration.
A range of tools offers big usability improvements for the complex network configuration challenges. Policy-based network management lets you capture broad descriptions or policies for what should occur in a network, including classifying and handling network traffic. New software components turn these policies into configuration instructions and deliver them to groups of network devices at the same time, in a coordinated manner. Figure 6-5 shows QoS policies being applied to network devices. QoS policies are interpreted by a policy server, which distributes configuration instructions to the devices in a network.
Figure 6-5. Policy Server Interprets QoS Policies and Distributes Configuration Instructions to Network Devices
With policy-based network management, you create a set of rules, called policies, that describe the behavior you want to see. For example, policies might describe which users and which applications are vital for your organization, such as VoIP, e-mail, ERP, and payroll applications. If congestion occurs, the network should give preferred treatment to traffic from these applications.
Policies that reflect QoS decisions cause configuration updates to be made in the network to implement the desired classification and handling of traffic. The policy server provides a central location for making automated, orchestrated device configuration changes to implement the policies you create.
Although applied policies lead to configuration changes in the middle of the network, the goal of these changes is to solve actual business problems by improving the end user's experience. Planning is critical. When developing policies, start by listing the problems you are actually trying to solve. Follow this with a list of possible solutions and their associated costs and risks. Determine who decides what solutions to implement, in what timeframe, and who writes the policies to guide the implementation.
Start your policy writing with a simple set of priorities: high, medium, and low. Decide which applications should be treated with high priority and which should be treated with medium priority, and then treat everything else as low-priority or best-effort traffic. Make the high-medium-low decisions for your mission-critical data applications. Then do the same for your voice and video applications, and decide how they should interact with your data applications. Avoid complex planning for day-of-the-week or time-of-day policies.
Policy-based network management is still evolving. Hardware and software from different manufacturers don't necessarily interoperate well, so it is hard to predict the effects of policies that result in QoS configuration changes. You will find that extensive testing and tuning are needed to understand the cause-and-effect relationships involved when deploying QoS. However, policy-based management should be a part of an overall plan to manage network performance and call quality. Its greatest benefit is that it helps to fit QoS into a larger, more cohesive, network performance management scheme.
Planning for Future Growth
Additional users and applications inevitably consume more network bandwidth and VoIP server resources. As more network bandwidth is consumed, you move from being overprovisioned (having plenty of resources) to being oversubscribed (having too few resources), a condition that can easily reduce call quality. As more VoIP server resources are consumed, servers can reach the limit of their processing capacity, which will also affect call quality and availability.
These management components will help you prepare for future growth. As you establish trends showing availability and call quality over time, you can tune your existing infrastructure and plan for future investments. With less IT time being spent checking error logs and tuning minute elements of your QoS solution, you will be able to focus more on real improvements to tighten and organize your VoIP system and management solutions.
As your VoIP system grows, more calls are placed on your IP network. Your system may grow from a single department to an entire site. The more users and sites that are in your VoIP network, the more important it becomes that the calls are being charged properly, to the right people and departments. You need to have good accounting and billing management to provide tracking and proper charges for the traffic as it grows.