Lorenzo Becchi

Orangepi 5 cluster

Created: July 7, 2023
Last update: August 6, 2023

I made a cluster, mini HPC, with some OrangePi 5 SBCs (Single Board Computers).

Reasons

A cluster of single board computers can be beneficial for a variety of reasons, depending on your specific needs and use cases:

  1. Increased Computational Power: By combining multiple single board computers in a cluster, you can achieve greater computational power than a single board can provide. This can be particularly useful for tasks that require intensive processing, such as data analysis, simulations, rendering, or AI/ML applications.
  2. Parallel Processing: Clustering allows you to divide tasks into smaller sub-tasks that can be processed in parallel by different boards simultaneously. This can significantly speed up the overall processing time for certain types of workloads.
  3. High Availability: Clustering provides redundancy, ensuring that if one board fails, the others can continue to work, thus increasing overall system reliability and availability.
  4. Scalability: As your computing needs grow, it’s often easier and more cost-effective to scale by adding more single board computers to the cluster rather than investing in a single, expensive high-end computer.
  5. Distributed Computing: Certain applications or algorithms are better suited for distributed computing, where multiple computers work together to solve complex problems or handle large datasets. Clusters enable such distributed computing environments.
  6. Load Balancing: Clusters can be configured to distribute workloads evenly among the boards, preventing any one board from becoming overwhelmed while others remain idle.
  7. Customization: Clusters allow you to create a customized computing environment tailored to your specific needs, whereas off-the-shelf high-end computers may not offer the same level of customization.
  8. Energy Efficiency: For certain workloads, a cluster of low-power single board computers might consume less energy than a single high-powered computer doing the same tasks, leading to cost savings and reduced environmental impact.
  9. Learning and Experimentation: Clusters provide a valuable learning platform for those interested in distributed computing, parallel processing, and other related technologies.
  10. Affordability: Single board computers are generally more affordable than high-end desktop or server-grade computers. Clustering these low-cost boards can provide cost-effective computing solutions for specific tasks.
  11. Data safety: Data safety is crucial for sensitive data, but using Cloud Services may not comply with local regulations. Processing locally with a scalable solution might be a good alternative.

Building

I started by ordering five OrangePi 5 and the cheapest cases I could find.

I asked for a M.2 2230 NVME SSD, 256GB, for each of them.

NVME M.2 disks on each OrangePi

I wanted to be able to pile up boards to have a small footprint.

Keeping the sides open make it for a better thermal dissipation.

Like that it looks quite pretty to me.

Less pretty once I added the network switch, the USB station, the fan and the cables.

I actually realized the I needed a better USB station because this one wasn’t offering enough power and the units would halt randomly.

Also I thought I would try to add an external USB fan to test several configuration with fan speed and energy consumption.

I know the rubber bands are everything but elegant.

I’ll stick to that for the moment. I might add more boards in the near future and I still don’t know the final size of it.

Specifications

Orange Pi 5

Costs

I tried to keep the prices as low as possible.

Buying the OrangePIs one by one made the price fluctuate on Ali Express, supposedly by the conversion between Yuan and Euros.

The total cost of the hardware: 834.78 Euros

Conclusions

Planning and assembling the cluster was technically easy.

The most challenging part was waiting for things to get delivered after ordering on Ali Express.