0

I am trying to test a technology stack that would utilize a cluster of parallella boards as data center appliances. My plan is to use Mesos with Docker, which seems entirely feasible. My question refers to the 16 core Epiphany chip on the Parallella itself. According to the documentation available from Adapteva, "Epiphany is best leveraged as a massively parallel autonomous application specific processor." So basically, do I need to write a custom tasking service that will allow Mesos to utilize the Epiphany cores, or can Mesos "discover" the Epiphany on it's own?

PS: I am aware that data center appliance use is not typical for a Parallella, but my particular use case requires a compact form factor and extremely low power consumption.

JBiggs
  • 101
  • 2
  • I highly doubt that this will work out of the box. According to the docs at https://www.parallella.org/board/ only the onboard ARM chip can "run" Linux, so I guess Mesos will not be able to utilize the other chip. – Tobi May 20 '16 at 19:12
  • So as I was thinking, I would have to write a service that would allow Mesos to discover the Epiphany, right? – JBiggs May 23 '16 at 12:16
  • IMHO, you'd have to extend Mesos itself, so that it can potentially use the chip's resources. Similar to hat is/was done with GPU resources (see https://issues.apache.org/jira/browse/MESOS-4424). I think it's highly unlikely that you can implement this in little time... Also, this will probably never be part of the official Mesos distribution. – Tobi May 23 '16 at 14:42

0 Answers0