7SIGNAL The Air and the Endpoint are Part of the System

7SIGNAL argued that the air and the endpoint belong in the network engineer’s definition of the system rather than in separate silos. The session introduced its architecture of device agents and Wi-Fi 7 sensors, the troubleshooting patterns they reveal, data collection and retention practices, and Iris, the company’s AI engine for root-cause analysis and remediation recommendations. Eric Cammuli opened the session at Networking Field Day 41 with Joe Tennant and Marko Pirinen, who founded 7SIGNAL in Helsinki in 2007.

Pirinen contrasted easy problems, such as falling beacon availability or a failed primary DNS server, with hard ones like business-hours slowdowns that require ruling out the radio before tracing the path to the application. He recommended keeping transmit power within three to five dB and stretching RRM channel changes to 12 to 24 hours. The sensor acts as a high-performance client that continuously joins the network to baseline it and catch DHCP, DNS, and RADIUS failures. The lightweight agent, running on Windows, macOS, Linux, and Android devices including Zebra scanners, uses about 1% of memory, collects scan data every minute, and runs short throughput, latency, and jitter tests for roaming, coverage, congestion, and interference analysis. Tennant said raw data is retained for 90 days in the US or EU and encrypted at rest and in transit, though customer-managed keys aren’t supported.

Tennant described Iris as a mix of statistics, procedural code, workflows, and multiple LLMs that triages issues, draws on a library of root causes, and returns next steps that can flow into a ticket. Because customers load their own service levels, Iris judges compliance against each organization’s definition of good performance. Iris is read-only by policy for now, though the team said MCP-based workflows can carry out changes and controller integrations with more vendors are coming. Cammuli argued that nearly 1,000 metrics once seemed excessive but now give AI the depth it needs, combining what the infrastructure sees with what the air and device actually experience.