Heroix explores new territory with Longitude

When it comes to deploying an enterprise-level systems management solution, one of the thorniest details is often maintenance of the ubiquitous data-collection agents. Depending on the size of the organization, maintaining the agent-related components of the enterprise software stack can become a real burden, especially when you factor in the need for revision testing -- to weed out bugs or incompatibilities with the agent processes.

Enter Longitude 1.0 from Heroix. A first of its kind, this systems management solution eliminates the agent portion of the equation altogether. In its place is a network of hublike statistics servers that use remote management protocols -- WMI (Windows Management Instrumentation), SNMP, rexec/SSH on Unix/Linux -- to collect metrics data over the network and aggregate it to a centralized data store (SAP repository).

Remote systems management is not a new idea. The basic protocols and infrastructure components have been in place for years. What makes Longitude different, however, is that it uses these facilities exclusively, as opposed to most traditional management solutions, which rely on a mixture of agents and remote sampling. This makes the product ridiculously easy to deploy: Just type in the name of the system or device you’re attempting to monitor on the Longitude Web console, and the product’s statistics server adds it to its list of sampling tasks.

Longitude supports a variety of remote authentication techniques and can collect data both from the host OS (Windows, Unix, Linux, and so on) and common app stacks (Apache, IIS, WebSphere, Oracle, and the like). These pre-configured rules profiles make setting up a basic monitoring infrastructure easy. They can also prove limiting, given that the list of supported applications is rather short and there’s no facility for defining your own custom profiles. Heroix promises to provide such a feature in the next version.

Even if your current platforms are supported, you’re still limited in how much you can customize the rules. For example, although you can tune individual threshold values -- to shape how data is analyzed and to determine when alerts should be logged -- you still can’t change the rules profile to include or exclude specific counters that might not be covered.

Another, deeper flaw involves the data-collection architecture. If Longitude’s greatest strength is its capability of collecting data without an agent, it’s also its biggest weakness. That’s because the collection process is entirely dependent on a persistent connection to the monitored entity. If network service is interrupted for any reason, Longitude can miss one of the scheduled sampling windows. And because the collection model is stateless -- each sample is a snapshot in time and unrelated to those that come before or after it -- there’s no way to recover those lost data points.

The ramifications for mission-critical environments -- where a missed event can mean the difference between detecting a potential failure mode in time to address it and sorting through the carnage postmortem -- are enormous. It also means that occasionally connected users -- mobile professionals and most home-office workers, for example -- are not viable candidates for instrumentation through Longitude.