The biggest technical challenge in transitioning from traditional circuit-switched voice and video systems to the new, more economical voice and video over IP packet-switched technologies is obtaining adequate quality of service (QoS) over the wide-area network. Quality of service is the capability built into the network to guarantee that information traverses the network without interruption, in a timely manner. Most existing data networks were designed for bursty applications that are not delay-sensitive, meaning that if a data packet arrives within a reasonable amount of time, both the application and the user are satisfied.
Voice and video data, on the other hand, are very sensitive to delay; if a packet arrives more than approximately 200 milliseconds (ms) after it is transmitted, the packet is worthless as a carrier of real-time communication because it will arrive too late to be used in the conversation or video image. Consequently, networks carrying IP voice and video must be designed and configured properly to ensure that real-time packets traverse the network efficiently.
The Battle For Bandwidth
The challenge of achieving adequate quality of service is exacerbated when a data packet must traverse the WAN. Typical local-area networks (LANs) run at 10 Mbps, 100 Mbps, and some even have bandwidth of a gigabit per second (1,000 Mbps) and higher. However, because bandwidth over the WAN is significantly more expensive than over the LAN, many wide-area networks operate at T1 speeds (1.45 Mbps) and slower, creating a huge bottleneck at the LAN/WAN interface. For normal data packets like e-mail, Web browsing, client-server programs, and a host of other applications, this LAN/WAN bottleneck is a nuisance, but not a performance killer because these applications can withstand delay and still function satisfactorily. However, when voice packets must compete with regular data packets for transmission over a bandwidth-constrained WAN, the voice and video applications may be rendered useless unless steps are taken to insure voice and video QoS.
All VoIP Traffic Is Not Created Equal
But how do you initiate the process of establishing a coherent QoS strategy for VoIP? The adage, you can only control what you can see is entirely applicable to this challenge. The prudent enterprise needs to identify which applications are currently running on their network, when they run, how much bandwidth they use, and if they are performing satisfactorily. A variety of tools, including packet sniffers and network probes, can be used to initiate a VoIP readiness assessment. However, relying only on low-level protocol analyzers, such as those that look at IP packet headers to classify traffic, gives no indication of which applications are running across the network and may preclude the discovery of significant traffic trends that can impact the performance and integrity of VoIP services dramatically. By deploying monitoring systems that analyze traffic at the application level (Layer 7), IT can determine the behavior characteristics of individual applications and how that behavior influences WAN ap plication performance.
Rich traffic classification is crucial you cant assess or control an applications performance if you cant distinguish its traffic. For example, P2P, viruses, and worms hide in an HTTP tunnel using port 80 to traverse the firewall. The growing complexities associated with network traffic make sophisticated classification techniques a necessity. Simple IP address or static port schemes fall short. Classification must detect dynamic and migrating port assignments, differentiate applications using the same port, and use Layer 7 application indicators to identify applications.
Armed with this understanding, enterprises must protect VoIP, not only from data traffic, but from unsanctioned VoIP applications arising from instant messaging (IM) programs such as AOL IM, Yahoo!, and MSN Messenger. Furthermore, there is a distinct possibility that future P2P applications may start masquerading as VoIP traffic in order to gain passage through NATs and firewalls and to receive priority treatment when traversing the WAN.
Controlling VoIP Performance
Once VoIP business applications are accurately identified, the next step in optimizing VoIP is to create a reporting methodology that examines three primary variables:
Performance: how long application traffic spends in different portions of the network.
Utilization: bandwidth usage by applications, locations, and users.
Diagnostic Aids: Information that helps when trying to analyze and pinpoint a problem.
IT must ensure that it does not succumb to simply collecting data. To manage VoIP applications effectively, comprehensive reports that succinctly analyze bandwidth usage, response times, the impact of configuration changes, and sources of delay, broken down by time spent on the network and server, are crucial. Since each network is unique, reporting tools must have the flexibility to measure, graph, and/or export numerous metrics (in some cases exceeding 100), in order to accurately capture usage, availability, efficiency, response times, errors, and diagnostics.
Armed with the right reporting tools and data, IT is better prepared to address three critical performance issues that must be controlled to deliver VoIP QoS:
1. Latency the end-to-end delay in delivering the voice stream from the speakers mouth to the listeners ear;
2. Jitter the unpredictable, variable delays in the delivery of each voice packet; and
3. Packet Loss the dropping of individual packets caused by network congestion.
Each of these three issues can cause significant degradation in voice quality and overall system reliability. Because VoIP is real-time, two-way communication, it is very sensitive to delays in the network.
Acceptable VoIP quality requires a bi-directional latency or delay of not more than 80ms for true toll-quality voice communication. Voice quality degrades as latency increases, but even with a delay of 150-180ms each way, voice quality is still in the acceptable range. In addition to the voice stream itself, latency issues must also be addressed with other VoIP protocols (SIP, H.323, MGCP, etc.) that handle the call control functions between two systems or users. In fact, these signaling protocols are often even more sensitive to delays in the network.
Some network devices attempt to overcome this problem by employing various queuing techniques (DiffServ, 802.1 p/q, IP ToS, etc.) to ensure voice packets take priority over other traffic waiting to get on the network. This is helpful to a certain extent, but being first in line to get on a crowded freeway doesnt mean youll get to your destination quickly. Whats required is more stringent, intelligent bandwidth management/QoS that can classify all VoIP-related protocols, allocate a guaranteed amount of bandwidth to each traffic type, appropriately prioritize VoIP traffic as it traverses the WAN, and provide granular assurance of VoIP on a per-call basis.
If jitter causes consistent delays in excess of 2030ms, voice quality will suffer. Some VoIP vendors have tried to solve this problem by introducing their own jitter buffers or queues to temporarily store and smooth out the delivery of voice packets. Likewise, routers also offer queuing mechanisms for the same purpose. Both options, however, can exacerbate the problem by actually contributing to delays. Therefore, minimizing and controlling network jitter is required to prevent the disruption of VoIP traffic. And this calls for the ability to assign and maintain a guaranteed rate and quality of all voice and data traffic across the WAN, which are prerequisites to delivering VoIP QoS. These technologies are functionally similar to conventional queuing, in that they smooth delivery of the traffic, but do so in the context of a more intelligent, policy-based bandwidth management/QoS strategy.
Because IP is a best effort protocol, if left unattended it will always be subject to unpredictable performance, including packet loss. Like jitter and latency, packet loss can be very disruptive to VoIPs performance. This is usually not an issue in the corporate LAN and should not be a problem if an enterprise relies on an MPLS service across the carriers backbone network. But at the bandwidth-constrained LAN/WAN boundary, where there is much greater contention for space in a smaller pipe, congestion and packet loss can become a serious problem. Although packet loss of one percent or less is within the bounds of toll quality voice, once packet loss reaches three percent or more the listener will notice the conversation breaking up. Unless this problem is controlled, packet loss can lead to dropped calls and the possibility of VoIP system failure. The key to addressing packet loss is to apply more controls to the IP network, moving it from best effort to predictable, optimal performance for all bu siness-critical applications including VoIP. The value is delivered through a set of QoS and bandwidth management tools that minimize the congestion and unpredictability of IP and maximize application performance over the existing WAN, often without the need for costly bandwidth upgrades.
Only by applying a complete suite of technologies that include deep classification, detailed reporting, control mechanisms to proactively protect and ensure performance and the ability to apply compression gains to the benefit of VoIP applications is it possible to deliver VoIP QoS. IT
Jennifer Geisler is senior product marketing manager at Packeteer, Inc. For more information, please visit the company online at www.packeteer.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.