bk99.de entertain the web since 1997

E-waste turned router: OpenWrt and OPNsense on old x86 hardware

Summary

Maya Posch tries out an old Intel Atom PC as an open router and shows how driver choice, installation and reliable measurements decide success. The Intel D2500CC board uses an Atom processor with two 1.8 GHz cores and initially two gigabytes of DDR3. OpenWrt’s generic x86 image did contain e1000, but not the required kmod-e1000e package.

Ideas

  • Retired x86 hardware can replace proprietary home routers with open source systems you control.
  • OpenWrt stays lean but requires precise knowledge of the kernel modules needed on unusual hardware.
  • OPNsense bundles routing, firewall, traffic monitoring and extensions in a comprehensive management interface.
  • Missing network interfaces often point to the image and not to faulty hardware.
  • A system that installs is not yet a router proven suitable for continuous operation.
  • Used hardware extends its life cycle but can lose out ecologically through high power consumption.
  • Reproducible network tests need controlled counterparts instead of varying browser speed tests.

Insights

  • Open routers shift control, and with it responsibility, from the manufacturer to the operator.
  • Recycling hardware only pays off when reliability, energy requirements and maintainability convince together.
  • Driver packages form an invisible boundary between theoretical compatibility and hardware that is usable in practice.
  • For infrastructure, boringly stable operation is a more valuable result than spectacular peak values.

Quotes

  • Happy Routing – subheading by Maya Posch

Habits

  • Posch checks old hardware step by step with a boot attempt, interface check, client access and comparative measurements.

Facts

  • The Intel D2500CC board uses an Atom processor with two 1.8 GHz cores and initially two gigabytes of DDR3.
  • OpenWrt’s generic x86 image did contain e1000, but not the required kmod-e1000e package.
  • OPNsense was installed on a 128 gigabyte SATA SSD with ZFS and mostly default options.
  • The browser readings fluctuated strongly and did not allow a reliable comparison of throughput.

Critique

  • The experiment measures neither power consumption nor sustained load, packet loss or behaviour during updates and failures.
  • Public speed tests mix router performance with the browser, counterpart, provider route and current network load.

Remarks

  • The article documents an interim state; a systematic comparison of OpenWrt and OPNsense is still to come.
  • Thin clients or small firewall appliances can be more energy-efficient alternatives to the desktop case used.

References

Recommendations

  • Check the CPU architecture, network drivers and supported installation media before the conversion.
  • Measure throughput, latency, packet loss and CPU load reproducibly with iperf3.
  • Compare power consumption and purchase costs with an economical dedicated firewall appliance.
  • Back up configurations and test updates and recovery before using it in production.

Read the original article

Search the Web Archive