User-centric LAN testing--the new IT imperative: today's far more complex LAN environments require a different approach to measurement.
Meanwhile, the concept of the LAN grew, from simple bus, ring and bridged connectivity, to switched and routed systems. The phrase "local area network" expanded to include non-traditional network elements, such as firewalls, proxies, caches and VoIP gateways. When today's network administrators talk about their LANs, they use terms like virtual LANs (VLANs), network partitioning, quality of service (QoS), Web proxies, multicast, intrusion-detection-and-prevention systems, and firewalls-terms foreign to most network administrators 10 years ago.
With this rapid expansion in scope and hyper-evolution of technologies, testing the LAN today is no longer a case of just looking at the lower layers of the OSI model (Layers 1-3), but also focusing on what users actually experience at the upper layers (Layers 4-7), while also addressing overall system security. The trend in testing is moving from a network-centric view to a user-centric view. Some of the key drivers for the user-centric view come from advances in networking technology and drivers, including: virtualization, Webification, convergence and security.
Even moderately priced LAN switches today provide VLAN support and basic routing. As a result, large multitiered switched networks have replaced many of the older bridged and routed topologies.
PACKET TESTING NOW IMPORTANT
To effectively test the new topologies, network administrators have to perform logical reachability testing and VLAN leak testing. The new generation of LAN testers uses Layer 3 IP-based test tools that check logical reachability at Layer 3, and runs tests to ensure that broadcast domains and multicast groups are appropriately bounded by the VLAN partitions.
In addition, on higher-end LANs that incorporate QoS devices (often used to prioritize time-sensitive services over bulk data transfers), network administrators have to perform tests that involve the validation of IP packet latencies, throughput and packet loss for each of the services affected by the QoS policies. For instance, on a LAN that dictates that interactive traffic (telnet, Web) has precedence over FTP, network administrators should use tools that can mix different traffic types to demonstrate that interactive traffic does, in fact, get a better QoS (lower latencies, less losses) when the overall traffic load is increased. More sophisticated solutions will provide user-centric metrics, such as application throughput and response times, that better explain how the user experience changed as overall traffic load grows.
If virtualization and QoS can be thought of as changing a LAN's internal structure, Webification can be explained as expanding what a LAN is--the external structure. Webification is the movement of many of the traditional client-server services to a Web-based solution, such as human resource support systems, CRM systems, sales force automation systems and even things as basic as expense reports.
Some of these solutions are already taking the next leap, from Web front ends to Web services (XML- and HTTP-based distributed computing).
Users now think of the LAN as providing these services, so it is not just a network, but an entire system with two major interdependent parts: the network and the applications.
To effectively test the LAN and the services it now comes bundled with, network and IT administrators require solutions that not only look at the performance of the network itself, but can provide detailed metrics on how the combined system (network and Web application) is servicing its users. Users only care about how the combined system is working. For them, a network failure is a failure and an application failure is a failure. Users do not make the distinction, nor should they care. For network and IT testers, however, the tools they use need to allow independent testing of the network infrastructure, as well as entire system testing, all the while providing sufficient data to pinpoint where performance problems exist.
Another trend that is being given almost as much press as the Web is voice over IP (VoIP). VoIP in the enterprise is undeniably growing and brings with it different challenges on the LAN. To scale VoIP across an entire enterprise, the LAN must be able to provide low latencies and acceptable loss rates in the face of bursts of data traffic.
Prior to large corporate VoIP rollouts, network administrators need to use solutions that can simulate large amounts of VoIP calls, while mixing in realistic data traffic. While some Layer 3 IPbased test tools can simulate VoIP-like traffic, the ideal voice-resting solutions should provide some element of perceptual quality measurement, such as perceptual speech quality measurement, and perceptual evaluation of speech quality. This provides a more user-centric view of what is going on.
SECURITY NEEDS ADD CHALLENGES
Security also is a valid and growing concern within the corporate environment, and is being tightened to protect valuable data. Firewalls, intrusion-detection-and-prevention systems, and virus scanners are now integral to most local networks, and bring with them new challenges in LAN testing.
To adequately rest security, network administrators can turn to a new breed of tools that can simulate attacks (e.g., denial of service and e-mail with virus payloads) embedded within actual traffic. Furthermore, from a methodology perspective, security testing should be mixed in with performance testing, making sure that security devices can perform under load, providing good availability and security simultaneously.
The four key drivers--virtualization, Webification, convergence and security-dictate the technologies necessary to test the new generation of LANs. At the same time, the reporting metrics that network administrators need to focus on are those that users care about. Users do not ask questions like, "What's the packet loss rate on the LAN when it's 50% loaded?"
Rather, they want to know: Will engineering running high loads on its VLAN affect the finance department's database server response time? How long does that Web page take to load at 2 p.m. in the afternoon when 1,000 people are using the system? What is the voice quality on this VoIP call I am making? How many viruses get through detection when the LAN is under load?
Metrics in the early days involved frames per second, packets per second, total aggregate throughput, frame loss rates and packet latencies. Those are less relevant to users and management today, however, who are more concerned about transaction response times, application throughput, application latencies, transaction successes vs. failures and quality of a voice call. The network administrator must now devise test solutions that can provide reports based on these new metrics that make sense.
LAN testing today is more involved than it used to be, with new technologies, new services and new topologies. The bottom line, though, to LAN testing success, is remarkably simple: focus on the user. Network administrators who understand this and develop LAN testing methodologies that address real user problems and report metrics in user-understandable terms, will find themselves remarkably successful.
For more information from Spirent: www.rsleads.com/309cn-250
Testing the LAN today is no longer a case of just looking at the lower layers of the OSI model.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Testing & Monitoring|
|Date:||Sep 1, 2003|
|Previous Article:||Gigabit Ethernet transceiver.|
|Next Article:||How to design a wireless LAN: key issues: RF coverage sufficient capacity and accounting for RF signal-loss factors.|