Skip to main content

Managed Service Nightmare - Lessons Learned, Part I


By any measure, Bob Andreini's network was a nightmare. Even though he had outsourced the management of the network, downtime was measured in days. His team regularly mediated finger-pointing between international telecom providers. Even the configuration of his infrastructure left a lot to be desired.

Andreini, the global director of IS and IT at Measurement Specialties Inc., a $200 million manufacturer of sensors with 12 locations around the world, knew something had to change.

His company had grown through acquisition, adding sites in locations in Galway, Ireland; Versailles, France; Dortmund, Germany; and Bevaix, Switzerland, a village of 3,500 outside of Neuchatel. But this growth also meant an increasing patchwork of network connections.

Reliability was a challenge, especially with an IT staff of 35 and only ten of those people devoted to the infrastructure -- able to work only part-time on this key task.

Downtime Here Meant Downtime There
A hub-and-spoke infrastructure meant that all communications went through MSI's corporate headquarters in Hampton, Va., some by point-to-point connections, others through VPNs. "If we had trouble in Hampton, we had trouble across the whole network," says Andreini.

Part of the problem, he admits, came from the telecom provider's own growth problems. It didn't always have the best relationship with telecommunications providers in the countries where MSI did business.

"We ended up with higher communications costs, because they couldn't get good resell rates." And because it didn't have good relationships, where there were problems, there was a lot of finger-pointing. "We would eventually have to have our IT people figure out the problem," sighs Andreini. Then the guilty party would admit it was their problem, sometimes as much as two days later.

Unmanaged Services
Some of Andreini's team even had to spend time learning about Cisco router configurations because their provider, which was supposed to be providing maintenance, had outsourced it to yet another company.

Solving MSI's network management problems took multiple steps, including signing on with another vendor of managed network services. MSI worked with the new company, Virtela, a Greenwood Village, Colo.-based company, and its team to create a new infrastructure. Together they took a close look at what kind of infrastructure would serve the global firm best.

In part II: how MSI solved its managed service problems.

Popular posts from this blog

Why the Future of AI is Agentic but Precarious

We have now entered the AI Agentic era, according to the latest series of reports by Google's artificial intelligence (AI) researchers. The shift from passive generative AI models to autonomous AI agents that can plan, reason, and act on our behalf is the most profound digital transformation in decades. As  Applied-AI Initiatives replace deterministic code, a significant challenge has emerged. Building an AI agent is easy; however, trusting it is complex. The current AI market momentum reveals a stark last-mile gap. While a developer can spin up an AI prototype in minutes, roughly 80 percent of the effort required to reach production is consumed by the work of safety, validation, and infrastructure. The reason is simple: AI agents are non-deterministic. They can pass 100 unit tests but fail catastrophically in the field because of a flaw in their judgment, not a bug in the code. Core Architecture and the Problem-Solving Loop An Applied-AI agent is defined by the synergy of four co...