Managing a data network in 2005 is much simpler than in 1995 or even 1985. The tools are more intelligent, and the information available is more accurate and complete. But network managers have let this sophistication cloud over a fundamental trait of corporate networking in 2005 — application fluidity.
In the past, network managers focused on configuration and faults, the hot points where outage problems were diagnosed and resolved. Less demanding issues, such as performance, were addressed by adding bandwidth. Applications that used the network tended to be transaction-based. Performance could be measured, and poor response time always could be blamed on the IT department rather than the network.
Applications today can be a complex mix of data, voice and video traffic, all masked by the fact that they use IP. A corporate network must accommodate real-time voice telephony, instant messaging, video teleconferencing, file transfers, storage backup/recovery and peer-to-peer interaction, in addition to the corporate core-application transaction traffic. Application additions and deletions are fluid in nature and can occur almost instantly. Applications that utilize networks are becoming increasingly intelligent, using sophisticated middleware to enable direct application-to-application communication. This fluid state can only become more volatile with the construction of applications using a service-oriented architecture (SOA) and/or grid technology.
Network performance is again on the front burner. The adage, “If it’s not broken, don’t fix it,” will no longer explain out-of-date revision levels for software in routers and switches, and delays in converting a network from IPv4 to IPv6, which may be required to accommodate new application, server, storage and user demands.
The first change that must occur is a mind-set update: Network managers must realize that performance is an issue even if a network is operating without user complaints. Next, they must evaluate and begin to use a new category of test/monitoring/management software that will allow for a readiness assessment of a network before applications are introduced. This is especially true for a VoIP application. This new type of management software is application-aware and will look at a network and infrastructure components to identify problem areas before an application is deployed. VoIP is just one example. The same type of software can be used to identify offsite backup and recovery or any other application-specific network performance problems.
Finally, a change is needed in the way network performance management is handled. Performance now must be monitored and managed using a set of predetermined and agreed-upon metrics. Policy must be established and then translated into network monitoring criteria. In some cases, existing network management system tools can be used to perform monitoring tasks. In other cases, new-application intelligent software must be integrated into the management environment.
This type of software usually has four components — local agent, data gathering, analysis/report generation and repair/correction. A local agent may reside in a client, server, storage and even application software itself. An agent is the key element required to generate and monitor performance metric information, which is then gathered for real-time or future analysis. Finally, manual or autonomic actions can be taken. Today, in almost all cases, manual action is taken after a level of management approval. In the future, software itself can make changes required to meet performance expectations. With new carrier options such as dynamic bandwidth allocation and component update/upgrade technology, which promises zero downtime, a high degree of automatic performance management can be achieved in networks.
Always remember: Applications-aware management must go hand-to-hand with applications-aware networking.
–Dzubeck is president of Communications Network Architects, an industry analysis firm in Washington, D.C. He can be reached firstname.lastname@example.org.