Skip to content

How ARAM works in your rack

ARAM reads each vendor’s interface, acts within your policies, and records everything.

Deploys as
One container image
Reads
Redfish, DCGM, ROCm, Modbus, gNMI and SNMP
Reports through
A Redfish-compliant API
Sits under
Your DCIM, not instead of it

Connectors

One connector per interface. The vendors below are the ones each connector is built for today.

ComponentVendors and platformsInterface
BMCsOpenBMC, Dell iDRAC, HPE iLO, Supermicro, NVIDIA, AMDRedfish
GPUsNVIDIA, AMDDCGM, ROCm
CDUsVertiv, Schneider Electric, CoolIT, and CDUs that follow DMTF’s Redfish profile for CDUsModbus, Redfish
SwitchesSONiC, and fabrics that expose gNMI or SNMPgNMI, SNMP

Each connector is built to the vendor’s published interface and tested against simulators and emulators. A pilot is where ARAM is validated on your hardware.

What it watches

Each component becomes a typed resource in ARAM’s semantic graph, so a fault in one domain can be traced to its cause in another.

Compute

  • GPU
  • Memory
  • NVLink
  • UALink

Power

  • PSU
  • Power cap

Cooling

  • Fan
  • CDU
  • Coolant loop
  • Leak

Network

  • Switch
  • Fabric

How it works inside

Telemetry moves through six stages, from the vendor’s interface to an action on the record.

  1. Connectors

    Read each component through its vendor’s own interface.

  2. Semantic graph

    Turn every component into a typed resource in one model of the rack.

  3. Models

    Detect faults, predict failures, and trace each one to its root cause.

  4. Policy engine

    Check each action against your policies: run it, wait for approval, or recommend.

  5. Runbooks

    Act, confirm the fix with consecutive clean readings, or escalate.

  6. API and record

    Report rack health and every action, with its evidence and policy.

Autonomy at your pace

You decide how far ARAM goes on its own, policy by policy.

  1. Recommend

    ARAM detects and diagnoses, then tells you what it would do.

  2. Approve

    ARAM proposes each action, and a person approves it before it runs.

  3. Automate

    The policies you trust run on their own. Everything else still waits for you.

See what ARAM does on your racks

Racks you choose, vendors you already run, and policies that decide what ARAM may fix on its own.

Request a pilot