Network Computingwww.networkcomputing.com

VitalSigns, on the other hand, passively monitors actual transactions at the client, recording performance and failures to a centralized server, where the VitalAnalysis application tracks adherence to service levels. VitalSigns also provides easy-to-understand graphical feedback.

How's the App? Logic would seem to dictate that the best approach to service management would encompass all perspectives and disciplines. It's a nice theory, but for manageability's sake, it pays to focus on one management discipline. This strategy may net you sufficient service information, while limiting the cost of acquisition and deployment.

Application performance is often the best element on which to focus, because it's most meaningful to the end user. Accordingly, many products target performance measurement. The four basic approaches to application tracking are API, transactional, database and network.

Tivoli and HP are the major proponents of the API method, with their Application Response Measurement (ARM). The idea is to hook performance characteristics into the application to monitor and diagnose its performance. However, this necessitates special coding and creates some overhead to the applications. To complicate matters, Microsoft is not expected to support ARM in its applications or tools.

Microsoft and Optimal Networks both prefer a transactional model for measuring application performance. Microsoft's Visual Studio includes a tracer for debugging code, but it is limited in that it can relate only an application's coding efficiencies.

Optimal Networks Corp.'s Application Expert takes a more comprehensive approach, using Network Associates' Sniffer trace files to build an off-line model of network application performance. Optimal includes the amount of time spent on network bandwidth, network latency, client turnaround and all servers traversed by a particular transaction. The resulting graphical map lets you raise and decrease server response and network bandwidth to predict achievable transaction response.

Products from Platinum Technology, Technically Elite and NetScout Systems use wire sniffing to feed application-performance measurements. Platinum's WireTap and TransTracker analyze the network for protocol and database transaction usage with software proprietary probes. NetScout and Technically Elite have extended their RMON probe technology and are creating application modules that recognize off-the-shelf software from SAP and PeopleSoft. The challenge in both cases is the strategic placement of the probes, coupled with their ability to maintain high accuracy when under load. Implementation requirements are limited to probe placement, and no additional processing is placed on the servers or clients.

Concord Communications' Network Health gathers data from SNMP MIB II and RMON agents reporting on network usage. The reports offer extremely granular detail on network reliability. Missing are modules to get at the application brass ring. Still, the level of detail and totality of the reporting make it a strong tool for eliminating the network as an application delivery hurdle.

Another approach to application truth comes from vendors that attempt to measure database and server performance. Database application monitors--such as Knowledge Modules in BMC Software's Patrol--take advantage of special knowledge of Oracle, PeopleSoft and SAP applications. And BMC can reach below the application into the database to diagnose a problem.

Like BMC, Empirical Software focuses on monitoring service levels at and on the server. Empirical drills down into very specific performance matrices of server OS, database, CPU and memory usage. It creates a health index that categorizes the state of services. This server-centric view tracks Oracle financial and manufacturing applications, as well as some PeopleSoft and Baan transactions.

Bridging the gap between network and systems management products is Micromuse USA's Netcool OMNIbus. It gathers and correlates data across network management, systems management, enterprise management, SNMP, PBXs, NT and just about anything that can create a log file--all without any configuration, providing value right out of the box. It offers a single screen of widely correlated data providing monitoring and diagnoses of service levels. Correlation of events through so wide a vision obviously monitors across all points of view, but also provides diagnoses to a component level.

International Communications Software's Continuity is tightly integrated with Cabletron Systems' Spectrum to extend the latter's alarm mechanism. It ably monitors SNMP data and other Spectrum-integrated data sources such as BMC Patrol. On its own, Spectrum has a complex alarm hierarchy based on network port-level connectivity, but Continuity extends the alarms so Boolean logic makes for worst- and best-case service monitoring.

Send your comments on this article to Bruce Boardman at bboardman@nwc.com.