South Florida (multiple businesses) · Past field engagements; exact dates not published
Case Study: A Custom Go Agent Watching Point of Sale Fleets in South Florida
Point of sale terminals fail in the worst possible way: silently, and in front of a customer. Businesses across South Florida, in the Miami area, Fort Lauderdale, Wellington, and beyond, were running registers with no health visibility at all; the first sign of trouble was a line forming. Commercial monitoring suites existed, but their per-node licensing priced out exactly the small businesses that needed them. So KYTHEL wrote one: a native agent in Go, purpose built for the job.
By Lazaro Anias, founder & engineer. Published . Updated .
- Location
- South Florida (multiple businesses)
- Project period
- Past field engagements; exact dates not published
- Scope
- Custom register health checks and a central heartbeat API
Where it stands: The agent shipped to multiple businesses across South Florida. In the largest deployment it reported on roughly 40 registers, and the licensing bill for all of that telemetry was zero, because the software was written for the job instead of bought for it.
The challenge
- Register hardware failing with no warning and no telemetry, discovered only when a sale could not be rung up.
- Commercial monitoring priced per node, which multiplies badly across a fleet of registers.
- Old and underpowered terminal hardware that could not afford a heavy monitoring agent eating its memory.
What we built
A native Go agent built for small machines
A compiled Go service collected health readings with a small memory footprint, recalled as under 12MB in the deployment. Resource use depends on the host and workload.
A heartbeat API instead of polling
Each register sent a scheduled heartbeat to a central API. Missing reports identified a host or connectivity problem for investigation; an absent heartbeat alone does not identify the cause.
Telemetry that matters at a register
Disk health, memory, service state, and connectivity, the specific failure modes that stop a sale, reported centrally so problems were visible before the customer saw them.
One codebase, many deployments
The same agent served multiple businesses with different fleet sizes, configured per site rather than rewritten, keeping every deployment maintainable.
The results
- Registers reported the configured disk, memory, service and connectivity checks centrally.
- The largest single deployment covered roughly 40 registers.
- The custom software avoided recurring per-register monitoring-license fees. Development, hosting and maintenance still had costs.
Clients not named by policy. Deployments spanned multiple South Florida businesses; the register count is from the largest single deployment and is not a claim about every site.
Related services