I have done a lot of reading about rpi clusters. Don't worry I am not going to ask if they function as a single virtual computer. I understand that they can be used to divide up complex tasks or space out tasks in parallel programming. Basically I want to create a home automation system, weather station, and add on a couple other projects I may find cool in the future. I just want to know if I can outfit each rpi in the cluster to do a specific one of these functions. They all require different peripherals attached to each rpi and I am not sure how much control over these peripherals I will have from the head node. Any clarification would be appreciated. Thanks.

  • As Patrick's answer implies unless you do have a purpose for the cluster beyond what's mentioned in the question, it's totally unnecessary. I.e., if what you mean is "I want a home automation system and a weather station, but then to sometimes do experiments with clustering", fine, but if what you mean is just you want a bunch of networked pis in a home automation system online with a weather station then don't bother with clustering, it will not benefit those purposes. – goldilocks Jun 6 '16 at 12:09
  • @goldilocks I do not have any clustering based projects on hand right now. I was just more thinking that if the projects progress that one rpi may have problems being able to do all of the projects by itself. That is a lot of peripherals to connect to one rpi. I figured one could be the weather station and another could be for certain aspects of home automation, another for a Web server, etc. I just want to know how much control I will have over the tasks and peripherals from the head node. All the projects I see usually consume a rpi and I want to spread out the load between many. – Christopher Jun 6 '16 at 13:57
  • The reason "all the projects [you] see consume a pi" is because they are demos, and someone is not going to explain how to create a weather station and make comments about an (unrelated) webserver running on the same pi, or use pictures where an (unrelated) camera cable, etc., adds confusion. – goldilocks Jun 6 '16 at 14:09
  • The limits you are more likely to encounter have to do with power, I/O, or the number of peripherals of a particular sort that can physically be connected -- in this case, load balancing the work will not help because load balancing does not provide you with increased amperage or additional physical connections on one particular pi and will detract from I/O potential. A weather station, for example, will use very little processing power; if you could power and connect enough peripherals, I think a single pi (any model) could easily run half a dozen weather stations. – goldilocks Jun 6 '16 at 14:09
  • I use the same (single core B+) pi as a server of various sorts at the same time as it runs an FM transmitter and collects barometric/temp data, with an 8x8 LED attached controlled by an app which samples and reports system metrics (processor load, RAM usage, net I/O, etc) several times per second. That pi is rarely under significant load because none of those tasks are that taxing in relation to the SoC's potential. It is not a 16-bit, <50 Mhz microcontroller, like, e.g., an Arduino. It's a monster by comparison in this sense. – goldilocks Jun 6 '16 at 14:09

Depending on how you build your cluster, you can ssh individually to each of the nodes from the head node, or even ssh directly from your PC into each node.

In short, your cluster can still act like individual Pis. They can still do individual jobs. Do whatever you want with them.

The cluster-ness will only kick in when you need it (e.g. when you run a cluster-based distributed calculation thing, or when you fold proteins). Otherwise, they will simply act as a networked group of Pis. Depending on how you build your cluster

|improve this answer|||||
  • I see, so I would not have any control over the other nodes from the head node? They would just be a network and I would have to bounce around to each rpi and control it's specific job from that rpi specifically? I was hoping to be able to oversee all of the different aspects from the head node. Although even if I could I think that the GPIO ports remain the property of the individual pi. – Christopher Jun 6 '16 at 14:00
  • @Christopher Ummmmm... Basically. In a sense, you do have control from the head node. It's either via SSH or some other tool (yes, there are tools for those). As for the GPIO pins, you may be better off getting the attention of joan via chat since comments aren't for extended discussions. – PNDA Jun 6 '16 at 14:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.