Why is network quality of service (QoS) so important and how can it be properly measured? This look into QoS discusses how it should be measured, by developing an analogy between QoS measurements and physical-layer cable certification tests.
November 1, 2001
In the wire and cable industry, we tend to focus on standards-based certification tests, which assess the quality of the network’s physical layer. Most of us know what a Pass/Fail means when it comes to cable certification. This article will examine how to take the cable certification concept up through the networking layers. Can we guarantee a quantifiable level of performance for active networks? Equipped with the right test tool, can a cable installer verify the network’s performance? Let’s take a look at the meaning of QoS, and then try to answer these questions.
Network QoS is really nothing more than an indication of an end user’s ability to perform daily routines without interruption or delay. QoS measures the performance of real-life operations — for instance, the speed of a file backup or the time it takes a Web page to load. As businesses increase their dependence on IP networks for critical data, video conferencing or voice transmissions, they have to be sure they can deliver information reliably, even in conditions of heavy traffic and congestion. QoS monitoring provides quantitative measures of a network’s ability to meet these needs.
Unlike data, live voice and video transmissions must be sent continuously in real-time and without delay. To ensure the delivery of time-sensitive voice and video traffic, the potential of the network must be determined. The ability to support multimedia and telephony applications at specific performance levels can then be verified. If the network’s potential is not great enough to support these demanding applications, no amount of traffic prioritization or increased device bandwidth will guarantee the level of service a user requires.
PHYSICAL LAYER MEASUREMENTS
The basis of a multimedia-ready, high-performance network is high-performance cabling. The impact of the physical infrastructure on a network’s QoS is often overlooked, but cabling quality can have a dramatic effect. Cabling standards such as TIA-568-B and its forthcoming Category 6 addendum are designed to ensure that your cabling will support current and future communications technologies.
If the physical cabling in a building does not meet minimum standards, investments in sophisticated management applications or routing protocols will bring limited returns. Cabling certification to ensure compliance with Category 5e specifications is a necessary indicator that a network can provide the QoS required by today’s most demanding applications. Certification to Category 6 specifications indicates that a network is capable of providing an even higher QoS, which will translate into enhanced performance for emerging applications.
Basic cable testing and certification are essential to building a sound network infrastructure. The remaining steps in delivering verified QoS are to check the active network configuration, measure actual performance data and analyze this data to predict real-world performance.
To provide an end-to-end indication of a network’s service availability, physical test data must be incorporated with a comprehensive analysis of the network’s operation and topology. This can be accomplished through simple modifications to the tools and methods commonly used by cable installation technicians.
Before addressing the performance of the overall network, all of the devices on the network must be identified, and each must be accessible and properly configured.
While physical-layer errors can initiate a packet retransmission and quickly use up bandwidth margins, common set-up errors can also interfere dramatically with overall network performance. Bad subnet masks, misconfigured servers and duplicate IP addresses can all play havoc with network traffic patterns and cause critical devices to fail. These problems can be detected by correlating protocol errors with previously gathered performance data — signal strength, noise, cable length and fault locations. For example, all network communications must first negotiate a route using IP (Internet Protocol), the de facto standard for transmitting data over networks. The routers are responsible for determining how data finds its way from one host computer to another and must resolve domain names versus a domain name server. To resolve their own IP address, they negotiate with a Dynamic Host Configuration Protocol (DHCP) server. A physical-layer issue in one area of the network may impact communication to these key devices, which could disrupt [FM1]communication between an end user and a file server. Misconfiguration of a router or gateway could also make the device inaccessible, leading to the same result. Indeed, many network devices — including proxy servers, authentication servers and gateways — impact communications throughout the network. If a physical error or misconfiguration interferes with these systems, overall network performance will be compromised.
By certifying that the cabling quality is up to standards and verifying that the network topology is intact, reliable transmission to these critical devices can be guaranteed. The concepts and methods of cable certification are therefore applied to higher level networking functions. Packet loss, over-burdened servers, 10/100 switch port auto configuration errors and poorly implemented path selection can all be identified at this level of testing. Test instruments supporting these functions can lower operational costs and improve efficiency within IT departments by enabling technicians to provide more advanced services without additional training or specialization.
TAKING VERIFICATION HIGHER
Once the physical characteristics of a network have been measured and the network topology has been verified, other factors influencing network performance can be considered. The responsiveness of key network resources such as file, application and web servers, e-mail servers and printers can be measured. The results can be analyzed to ultimately provide a measure of real-world network performance: QoS.
When installation technicians can measure the responsiveness of specific devices at specific times — using tools previously reserved for complex systems and expert network integrators — they can certify that their networks are ready for advanced applications.
Initial QoS measurements can be used to develop quantitative guidelines for network performance. Technicians can establish a baseline against which future performance can be compared. When QoS data is collected and stored in a database, network administrators can identify utilization trends. As the network evolves over time, the historical data can be used to remove bottlenecks and to plan upgrades or reconfigurations.
ENABLING PERFORMANCE TESTING
With the previously described theory of performance measurement, a valuable and reliable network analysis is possible. With this type of analysis, complex network design decisions can be made, capacity planning is simplified and network integration is greatly facilitated. In addition, “front-line” technicians are able to test and rate network performance on their own and quickly resolve complex network issues.
To convert theory into reality, the advanced physical- and network-layer data must be collected and analyzed, which is no simple task. Fortunately, new test tools and procedures are available to perform this process automatically and eliminate human oversight and error.
Measurements that once required complex network management software and specialized equipment can now be performed by field-test technicians using inexpensive hand-held testers. These tools include capabilities such as network auto-discovery, which surveys the network to generate a complete list of workstations, switches, routers, servers and other network devices. The tools can be used to measure performance and availability of network services, and to identify and correct common network-layer misconfigurations such as subnet masks or duplicate IP addresses. The verification of network performance — which applies the methodology of cable certification to the active network — is a
new feature of the latest generation of field test equipment.
Adding real-world network testing features to cable test equipment allows network administrators and installers to verify that their networks will meet the performance requirements of bandwidth-hungry voice, video and high-speed data applications. These features correlate test and measurement data from the physical cabling up through the network protocol layers. Performance monitoring capabilities capture a picture of a network’s health, while troubleshooting functions make it possible to pinpoint problems quickly and easily.
In any organization, business-critical network services must be tested to ensure that the users can perform their jobs effectively. As noted, there can be a three-stop methodology for optimizing network performance: certifying the integrity of the physical layer; verifying the proper network and device configuration; and measuring QoS to ensure that the network is delivering the required level of service.
By elevating the concepts and methods of cable certification to higher network layers, users gain the ability to quantify network quality, allowing them to administer and manage the network at a much more complex level.
Philip Lippel is a Member of Technical Staff at Agilent Technologies’ WireScope Operations in Marlborough, MA. He represents Agilent on TIA and ISO standards committees and has been active in the development of the TIA-568-B and Category 6 standards.
Karen Kiernan is the Marketing Communications Manager at Agilent Technologies.