Technology

The technology behind AssureMesh.

A tenant-scoped data pipeline, a multilayer network digital twin, an evidence-first RCA engine and predictive models, engineered to work as one.

gNMINETCONFeAPISNMP v3LPM / FIBECMPmin-cutleast-squareshash-chained audit

01 / Data pipeline

Every signal, one tenant-scoped stream.

Device and cloud telemetry lands in a per-tenant pipeline, where deterministic detection turns raw signal into typed faults.

Sources

Cisco IOS-XEgNMI · NETCONF · SSH CLI
Arista EOSeAPI
Juniper JunosNETCONF
SNMP devicesv2c · v3
Public cloudAWS · Azure · Google Cloud

Transport

Edge collectorOn site, next to your devices
Cloud agentsOne per provider
Encrypted streamingInto the control plane

Ingest

Tenant-scoped ingestIsolated pipeline per tenant
Metric historyTime-series store
Network stateDevices, links, BGP, LLDP

Detect

Fault detectionInterfaces, CRC, flapping, duplex
Resource and routingCPU, memory, temperature, BGP
Burst suppressionExact de-duplication, correlation

02 / Digital twin engine

A versioned, multilayer model of the network.

The twin is a tenant-scoped multigraph with a real routing engine, so every answer about paths and impact is computed from the model itself.

Layers
Physical, routing and dependency, with coupling between them
Snapshots
Versioned, so any two states can be compared
Routing engine
Longest-prefix match and FIB, ECMP, k-shortest paths, dominators, min-cut
Answers
Reachability, isolation, paths, SPOF and redundancy, change verification

03 / RCA engine

Every finding cites its evidence.

Two stages. Deterministic patterns produce the diagnosis on their own, and optional AI enrichment adds a readable narrative on top.

Stage 1
Deterministic pattern matching over faults, topology and timing
Stage 2
Optional AI enrichment for narrative and context
Evidence
Interface state, counters, adjacency, routing and change events
Output
Likely root cause, blast radius and an RCA report draft
EVIDENCE CHAIN · EXAMPLE
E1Te1/0/48 oper-down at 09:41:07interface
E2CRC errors rising the hour beforecounters
E3LLDP adjacency to core-01 losttopology
E4branch-22 BGP neighbor idlerouting
STAGE 1Deterministic pattern matchCORE
STAGE 2AI enrichmentOPTIONAL
Finding · INC-2041Likely root cause: uplink Te1/0/48 on dist-sw-03
E1E2E3E4

04 / Prediction models

Forecasts with a confidence level.

Least-squares trend models fit each metric's recent history and project it forward, so the NOC sees what will cross its threshold and roughly when.

Model
Least-squares trend regression over metric history
Signals
Interface errors and CRC, utilisation, BGP stability, CPU, memory
Output
Projected trend, confidence level and time to threshold
TREND FORECAST · EXAMPLE
THRESHOLD NOW TIME TO THRESHOLD METRIC HISTORY

05 / Built to be verifiable

Proof built into every layer.

Every finding carries its evidence, every action lands on a tamper-evident record, and every release is signed.

Evidence on every finding

Root causes cite the interface state, counters, topology and routing events behind them.

Tamper-evident audit trail

A hash-chained audit log records every decision and every step.

Signed releases

Sigstore-signed builds with SLSA provenance you can verify yourself.

Isolated and read-only

Row-level tenant isolation, TOTP MFA, and read-only operation by default.

Read about our documented security programme.

Next step

See the engine on your network.

Tell us what you run today. We will walk you through AssureMesh on your kind of network.