
Your monitoring system, read-only.
Step two of the join: utilisation read from your monitoring system over one read-level key. PRTG is in progress; the others are planned.
Request a pilot
- What it does
- Once a monitoring system is connected, reads hourly and daily utilisation aggregates per interface, binds an interface to a service through a reviewed queue, and supplies the delivered megabits behind the metric.
- Status
- In progress. PRTG is tested against recorded fixtures; the reads and the binding queue are not yet in the product.
- What it reads
- Your network monitoring system (NMS) and its own API, on a read-level key. Interface inventory and historic series. PRTG is the first connector and is tested against recorded fixtures.
- What stays yours
- Your devices, your SNMP, your traps and your write access, exactly as they are. We work from one read-level key to the platform that already polls them, and a missing sample stays missing rather than being interpolated.
- Where it shows up
- Cost per megabit delivered, the utilisation pane, and the coverage gate that decides whether a finding may be raised at all.
Who uses it
- The reader
- Network operations, who own the monitoring system, and group finance, who need a denominator under the cost.
- The question
- How much traffic did each service actually carry, and how sure are we of that number?
- What they get
- Hourly and daily utilisation per service with the coverage of every bucket beside it, a binding queue that keeps interfaces on the right services, and a cost per megabit delivered that states how much of the month it observed.
- Where it starts
- One read-level API key from your network team, and confirmation of which interface belongs to which service.
How it is built
Works with your monitoring
Riopex reads the system your network team already runs, which is a smaller ask and a faster start than standing up polling infrastructure inside fifteen countries' networks. That path costs, in our estimate, six to twelve months of infosec review, penetration testing and data-processing agreements per customer, for data the customer already collects.
Coverage travels with every read
Coverage is the fraction of a bucket that actually had data. Zero coverage means missing, never zero bits per second. A twenty-minute hour is a one-third-coverage hour. Nothing is interpolated, and a gap suppresses findings rather than creating them.
A quiet link is a working link
The utilisation pane renders neutrally, with the circuit's role and a coverage chip beside the number. Backup, DR, contractual minimums, burst headroom and seasonal capacity are legitimately billed, and a design that colours a quiet DR link red loses the room.
Binding is a queue
An interface binds to a service automatically only on an exact identifier match. Everything else opens a work item with ranked candidates. Bindings are temporal, so a rebind creates a new period and never rewrites months of history.
What you get
A denominator with a confidence
Every utilisation figure carries the fraction of its period that had data. A quiet hour and a missing hour are different facts, and the figure says which it is.
The network left as it is
Your polling, your SNMP and your traps stay between your devices and the system that already collects them. Riopex reads that system's own aggregates, outbound, on a request budget your monitoring has first call on.
Findings that clear a bar
A low figure becomes a finding only after the circuit's role, its contract and its coverage have been checked. A backup link at two percent is a working link, and the pane says so.

How it works
- ConnectYour network team issues a read-level API key. The key is stored as an opaque reference into a secrets manager and never enters the database or the repository.
- DiscoverWe list the devices and interfaces the key can see, and propose bindings to services already in the inventory. You confirm them; we do not guess.
- ReadHourly and daily aggregates, which is exactly what a cost question needs. Requests are budgeted deliberately, because over-driving a monitoring core would degrade your own monitoring, and a slow first import is the better trade.
- NormaliseUnits are declared by the connector and normalisation happens in exactly one tested place. A rate source that returns counters raises an error, which is what keeps a factor of eight out of your cost per megabit delivered.
One circuit, end to end
How a month will run once PRTG is connected: a 200 Mbps dedicated internet access circuit at a branch, read across one month.
- The key
- Your network team issues a read-level PRTG key. It is stored as an opaque reference into a secrets manager and resolved only when a read runs.
- Discovery
- The connector lists the sensors the key can see. One interface carries the circuit reference in its alias, so it binds automatically; two others open review items with ranked candidates.
- The read
- Hourly and daily aggregates for the month are read, with PRTG's own coverage figure per bucket. Twelve hours in the month have no samples and are recorded as gaps, with coverage reduced accordingly.
- Normalisation
- PRTG reports a rate in bits per second. The connector declares it as a rate, and the fixture test that asserts the declaration is what keeps a factor of eight out of the figure.
- The figure
- Peak daily utilisation lands at 61 Mbps against 200 contracted, with 98% coverage. The circuit's role is primary, so the figure is eligible for review, and the cost side comes from the invoice line it is bound to.
Questions about bandwidth
Why aggregates rather than raw samples?
A cost question is answered by hourly and daily aggregates, and reading raw samples would put load on your monitoring core for no better answer. The monitoring page sets out the request budget.
What if an interface is monitored by two systems?
Each binding is temporal and names its source, so two readings of the same service are two rows, and connector policy decides which one supplies the denominator. A source derived from session logs, such as a firewall analyser, is capped by policy and stays below the bar on its own.
Does a quiet circuit count as waste?
A quiet circuit is a question, and its role answers it. Backup, DR, burst, seasonal, test and out-of-band management roles block a waste finding, and an unknown role blocks it too and becomes an onboarding task.
How is the key kept safe?
Read level only, held as a reference into a secrets manager, resolved at job time, and absent from the database, the repository and every environment file. The security page describes the rest of the control set.
Related capabilities
