Edge computing: software as close to the field as possible
Not everything has to go up to the cloud. A kiosk in a shop, a unit on a machine, a router on an isolated site must decide on the spot: within milliseconds, with a network that drops, over a 4G or satellite link paid by the megabyte. Edge computing means placing part of the software where things happen, and carrying to the cloud only what matters.
KERN-IT built this expertise from both sides: through research, with Kenobi and its 2,000 network points supervised across Europe, and through the daily operation of self-service kiosks that have been running for more than ten years, network or not.
Three situations where the edge changes everything
The unit in the shop
A copy service kiosk identifies the user, reads the USB stick, prints and counts on the spot. If the shop's network drops, it keeps serving and synchronises the till when the connection returns.
The isolated site
A site connected over 4G or satellite cannot send every measurement every second. The node filters, aggregates, keeps the history locally and reports the essential: that is what we do for the sites supervised by Venn Telecom.
The machine that sorts its own data
A connected sports machine produces thousands of measurements per session. The embedded electronics compute on the spot what matters (cycles, load, duration) and send only the result.
Kenobi, our edge R&D
In 2020, with the operator Venn Telecom, we launched Kenobi: 24 months of research co-funded by Innoviris and VLAIO to prototype a vendor-independent network management platform. The edge part of the project consisted in designing an edge computer connected to the SD-WAN equipment and experimenting with its remote control: deploying and updating nodes without site visits, containerising and orchestrating processing on the edge, deciding where intelligence sits between the cloud and the field depending on the latency, bandwidth and reliability of the link.
The project supervised more than 2,000 network points across Europe in real conditions, prototyped machine learning problem detection and gave birth to GISWAN, our tool for visualising network quality in the field. Its main lesson fits in one sentence: automation earns its place when it decides fast and locally, and a human stays necessary when the decision commits the customer. The full story is on the Kenobi project page.
What it changes for you
It works without a network
The service carries on during the outage; the data synchronises afterwards. Your figures stay complete.
Less data carried
Sorting happens on site. On a cellular or satellite link billed by volume, that is one line less on the invoice.
A fleet that updates itself
New software version, new configuration, new content: pushed remotely to every unit, with no travelling.
Decisions in milliseconds
Cut, alert, switch over, refuse: the decision is taken on the device, without waiting for a round trip to a server.
Frequently asked questions
The question to ask is "what must happen if the network goes down, and how fast must the decision be taken?". Whatever must carry on without a network or react in under a second is handled on the device; whatever needs an overview, history or a human is handled centrally. Most of our architectures combine both.
An off-the-shelf mini PC or single-board computer is enough in most cases, chosen to last and to be replaced easily. On network equipment, we use the manufacturer's gateway whenever possible. The hardware matters less than the recipe that allows it to be reconfigured and replaced in a few minutes.
Each unit is described by a versioned configuration. An update is pushed in waves (a pilot site, then the rest), with a check that every device has received and applied its version, and an automatic rollback if a unit does not restart properly.
It is a risk to manage, not a reason to give up. The unit talks only to the platform, over an encrypted channel; its ports and session are locked; it stores only what it needs; it is supervised, so its silence is detected. A well-designed device is safer than a workstation forgotten in a shop.
A pilot unit with its supervision platform starts at €20,000. A complete fleet (units, platform, remote updates, alerts, portal) ranges between €40,000 and €100,000 depending on the number of sites and integrations. See our price ranges.
A fleet of devices that must run without depending on the network?
Describe your sites, your equipment and your network links. We will tell you where to place the intelligence, and where not to.