The world of high performance computing (HPC) evolved these past years, especially after the appearance of the beowulf clusters which consist of commodity Equipment interconnected to harvest their computing power. This cost reduction, made HPC accessible to a large public and some companies built their own cluster and datacenters based on the Beowulf concept. However, the costs are still relatively high and the clusters still consume a lot of energy.
In this study, we aim to provide a low-cost, low-energy cluster based on ARM devices. This architecture (ARM) is known to be more energy-efficient than any other architecture. Also, we implement security-related algorithms and execute them on our cluster to test its efficiency. Since combining clustering and virtualization has proved to provide notable advantages, we deploy Docker, a new lightweight virtualization solution, on some ARM boards to test its impact on system performance.
The results have shown that the ARM based cluster can deliver good performance, given the right message passing environment and load-balancing strategy. As for virtualization support, Docker had little impact on the board's performance but is not yet adapted for the ARM architecture.
| Date | 27 Apr 2016 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Chamseddine Talhi (Supervisor) |
|---|
Zagdene, G. (Author),
Talhi (Supervisor),
27 Apr 2016Student thesis: Master's thesis › Master in Engineering: Engineering