Aegis did not appear overnight. It began years ago as wireless monitoring, because there was none worth the name: what existed showed a controller with a green tick and nothing about what the people behind the access points were experiencing. I kept building on that, next to my day job, and it grew into a platform for the whole network.
Where the standard was set
I spent years building and running a PCI-compliant monitoring stack for payment service providers. In that world a blind spot is not something you get round to next quarter. It is a compliance finding, with an auditor attached. Everything had to be observed, and you had to be able to prove it had been observed. That is a demanding school, and the expectation stays with you afterwards.
What a thousand devices teach you
Today I work as a technical architect, and part of my job is making sure a fleet of more than a thousand network devices is properly monitored: switches, wireless controllers, access points and routers across a lot of sites. At that size the questions change. You stop asking whether a device is up. You start asking which sites are degraded, why, and since when and how long. Collecting everything is easy. Showing the few things that mean someone has to act tonight is the hard part.
Why the existing packages did not fit
Every time I looked at the usual monitoring packages, the same things got in the way. Anyone who has run one of them will recognise the list.
- They are server tools with networking bolted on. Disks, agents, processes and storage volumes take the foreground, while the wireless controller with a dead telemetry subscription sits three menus deep.
- Putting a switch in maintenance is a search. Three menus to find the option, and by then the alerts have already gone out.
- Looking at one link from three days ago takes a detour. Find the switch among hundreds, click down three levels, pick the port, set the time window. By then you have forgotten why you looked.
- After-hours alerting is a project of its own. You know exactly which devices the on-call engineer should hear about at two in the morning: the core, and only when it is critical. Setting that up, with escalation rules, time windows and exceptions per device, takes weeks. So the on-call phone either stays silent or rings for everything.
- They still poll SNMP. SolarWinds fetches statistics every ten minutes by default, PRTG scans every minute. Since IOS-XE 16.10 a Catalyst switch can push its own state every ten seconds over gRPC, model-driven telemetry. Ask the big packages about it and the answer has been the same for years: it is on the roadmap.
- They need weeks of configuration before they tell you anything. Templates, discovery rules, threshold sets, tuning per device, and it starts decaying the moment the network changes.
- The dashboards look their age. They were designed when a NOC was a room with a projector, and they are unreadable on the laptop where the work actually happens now.
- And they charge for it. The large vendors ask top prices, and count per sensor or per access point, so the bill grows with every device you connect.
That is not a criticism of those products for what they were built to do. They were built for something else, and no amount of extra configuration closes that gap.
What Aegis does instead
It starts from the network: sites, switches, wireless, and what the people using them actually notice. Telemetry works from the first day: onboard a switch and it streams its own state. Maintenance is one click. For the on-call engineer you mark your core devices, choose criticals, and the rest waits until morning; alerts go through SIGNL4 to whoever is on duty, with escalation and without noise. That link from three days ago: type part of the hostname, click the port, slide the time window back. It ships with opinions, so it is useful the same afternoon you deploy it instead of after a quarter of tuning, and it is built to be read at a glance, because the moment you need it most is the moment you have the least patience for a maze.
It runs on-prem, on your own VMware or Proxmox. One licence per monitored switch, router or controller; access points do not count, however many you have.
About how long it took
Years. The wireless monitoring came first, the switching, telemetry and site views followed, and every part was shaped by the audits and by the thousand devices. To be straight about the last stretch: AI shortened the distance between deciding something should exist and having it run. It did not decide what to build. That is why a platform built by one engineer behaves like one built by a team.
Questions, or want to see it against your own network? Mail me at [email protected].