Embedded cluster platform for a remote parallel programming lab

Single-board computers have recently grown to offer developers a wide range of options where the common denominators are low power and low cost. In this paper, we present an embedded cluster platform for a remote parallel programming lab to be used in an online course. A remote lab server handles all requests coming from the front-end running on an online learning platform and controls the execution of the parallel programming assignments submitted by students. The embedded cluster where the jobs run is made out of single-board computers connected through a gigabit network among them and to the lab server. In our first working prototype, we have tested six different state-of-the-art single-board computers, evaluating their processing latency, price, and tools compatibility. We found that the Vim3Pro performed best overall, being the fastest in most tests, having a mid-range price, and being only two times slower than a much more expensive high-end Xeon processor when using the same amount of cores.