A question of application

From NetworkComputing Magazine Vol 17 Issue 01 - January/February 2008

Douglas Smith, President of Network Instruments makes the case for application performance analysis on the network

Ensuring timely access to business-critical application servers is what justifies having a network (and an IT staff) in the first place. Effectively managing application performance means measuring how well key components of the IT infrastructure are performing over time, and identifying and efficiently correcting deficiencies before they adversely affect your business.
Network analysers with in-depth application analysis capabilities are crucial for defining network performance benchmarks, ensuring smooth application delivery and the successful roll-out of new applications.

Probe Deployment Decisions
An important consideration in monitoring with an analyser is probe placement. Most all-purpose network analysers are distributed, allowing a single console to monitor multiple switch segments via probes. Deciding where to place probes for application visibility in such an environment requires some thought and planning. A probe can only "see" traffic that passes through the network segment where the probe resides. In addition, because probes perform the time-stamping function that makes analysis possible, the location of the probe can itself affect response-time analysis results.
If the probes are placed at the edge of the network - for example, by monitoring SPAN (Switched Port Analyser - sessions obtained from switches at the edge), the network manager gets a more accurate reading of delay, rather like that experienced by users, but will need many more probes to monitor all of the transactions occurring across a multi-segment network.
However, by placing probes closer to the application servers (for example, by monitoring a SPAN session of the server ports on the core switch) the network manager can see all of the transactions, with fewer probes. While this may minimally mask how clients experience delays, the network manager will be able to determine changes in delay and identify important performance trends; all helpful in delivering a better user experience.
The benefit of full visibility achieved by placing probes closer to application servers is demonstrated through the following example. XYZ Company has a number of different servers attached to its core switch. Customer records and order processing depend on two SQL servers connected to the core switch: D1 and D2. The performance of these servers is monitored continuously by a probe appliance for 10/100/1000 networks, via a SPAN session configured on the core switch that mirrors all traffic flowing to and from D1 and D2.
With this setup, the administrator can track every single transaction, no matter where the client resides, and reduce the number of probes needed for monitoring.

Benchmarking and Continuous Monitoring
Once the analysers and probes are in place, monitoring of network and application performance can begin. Daily monitoring and long-term trend analysis are critical for efficiently managing application service delivery. When IT professionals monitor daily and study long-term performance trends, they learn how servers and the network behave when things are running at their 'normal'.
Application and network performance benchmarks can then be established based upon what is expected. When performance rises above or falls below the target level, alerts can be set to notify the appropriate person of potential performance issues.

New Application Rollout
With talk of implementing new applications on the network from VoIP to MPLS, and new CRM software always present, continuous network analysis is a must for ensuring successful application rollouts. Understanding overall bandwidth demand and application performance, and establishing benchmarks for acceptable network performance are critical for determining how the network will handle new application traffic and identifying the changes necessary to effectively support application delivery.
Communication applications like VoIP are sensitive to any performance degradation or network contention; planning and monitoring helps to ensure smooth performance. Network professionals must have the right analysis tools to ensure they can control delivery of business-critical applications.
Only with these tools can managers identify issues before they impact the user and plan appropriately for future growth. NC