Concord Communications' Spectrum Service Manager 7.1

Spectrum Setup

Running Spectrum Service Manager in our Syracuse University Real-World Labs®, I discovered hundreds of items, including switches, routers, Web and WebSphere servers, and services such as DNS and Mail. I used these to create service groups.

One of the first services I set up monitored the canned WebSphere application--I called it "Plant Store," because WebSphere's Plant Store was at the heart of the group. My Plant Store service included the WebSphere service, DNS, Web server and the switch to which the WebSphere server was attached.

Next, I needed to create a policy around my Plant Store service. From polices, you define thresholds as well as the impact a threshold violation has on the overall service.

Spectrum Service Manager comes with predefined policies you can customize by changing Value Maps and Rule Sets, the two basic sets of controls. Value Maps compare a condition to the defined threshold, check status or measure response time to determine if a service, server or device--whatever is defined in the service group--is running correctly. Rule Sets are configured to use the results of Value Maps; they notify the appropriate administrator if a service is up, down or degraded.

You can choose to send notice of service status from a single check-box option in Service Manager. You also can define or alter levels of impact, from low to high, from Service Manager. I created a policy for my Plant Store service that would send me a "critical error" notification if any part of the service stopped running. This meant that my critical error notification by default showed up in Spectrum's event-management screen as a red critical event because of the impact value I assigned. Being able to assign event status and define the impact each event type triggers is what really matters when you're monitoring services.Recycled Policies

Spectrum Service Manager makes use of service monitors already set up through nesting and reuse. For example, I defined a DNS service for redundant DNS servers. The policy rule showed that both servers had to be down for the service to be down; if only one was down, the service wasn't degraded, but a warning was indicated. This sophisticated piece of service logic would be time-consuming to re-create for all monitored application services that use redundant DNS.

But because the product lets you reuse created services in other services I only needed to define the redundant DNS service one time in one place to use it again with other applications that rely on redundant DNS services. Additionally, any changes the DNS service may require can be made in one place and then the changes will be replicated automatically.

Bumps in the Road

Populating the DNS service with DNS servers is a manual, static process requiring each DNS server to be defined individually. It would be better if service groups could be dynamically populated using information Spectrum discovers. Spectrum knows when a server is providing a DNS service--or any service, for that matter--so it would be simple enough to include in the service-grouping mechanism a query that would dynamically populate the group with the current appropriate service providing devices--DNS, in this example. Concord officials say the company plans on adding this functionality in the future.Upstream Rules

To make use of the Spectrum Layer 2 topology knowledge, the switch to which the WebSphere server is connected is also in the service definitions, but I didn't have to define any upstream rules about the switch and its relationship to the network. When I failed an upstream switch between Spectrum and the WebSphere server, Spectrum not only knew how the switches were connected (because the edge switch is in my WebSphere service), it also discovered that the root cause for the service being unavailable was the upstream network failure. This is radically different from most service monitors, which would understand the service was down but could not provide a root cause.

Spectrum can monitor a huge list of standard, enterprise and proprietary agents, the latter including Concord SystemEdge and Compaq Insight agents. There is also Concord's Spectrum iAgent, which offers Host MIB and proprietary monitoring. I installed iAgent on our WebShare service without a hitch.

Currently, Spectrum Service Manager can't monitor instanced attributes like CPU or disk partitions. Concord offers a work-around, however. It monitors CPU as part of its network device and server polling and will send an alarm on threshold violations. I set a threshold in Spectrum and then a policy in Service Manager that read the alarm as a service degradation rather than an outage. Concord says it plans to add service polices for instanced attributes in the next release, but this work-around was easy enough.

When a number of IT infrastructure pieces are needed to bring a service to end users, service management is difficult. Concord's Spectrum Service Manager 7.1 does a good job of reducing the amount of time involved.Bruce Boardman, executive editor of Network Computing, tests and writes about network and systems management. Write to him at [email protected].