What's Your Pi Plan??

My plan is to use RasPi for ... everything.

It's the first Linux device with Ethernet at a price that makes it viable to "buy a ton of them, use one per task".

Among the candidate applications:

- Home automation, networked controls and sensors on everything from lights to appliances (currently use X10).

- Home security, all PIRs, door and window triggers and the alarms networked and calling out to cellphone.

- Home entertainment, with the usual XBMC or MythTV in the living room and slave nodes around the house.

- High level access/monitoring of 3D printer (low-level realtime control is done by a dedicated microcontroller).

- Thin client workstations in all rooms, linked to the main home computer systems.

- Voice recognition and control, recognized commands routed to wherever is appropriate over the network.

- Servers for many different things, one RasPi per application:  for example a wiki for my cooking recipes.

- Experimental RasPi clusters.  I have a background in parallelism and concurrency, this is going to be fun.

- Music making, controlling my many dumb MIDI devices and giving them a level of musical intelligence.

- Open gaming system.  It's early days in this scene, but with millions of RasPi around, the day will come. :-)

And so on ...

Morgaine Dinova wrote:

 

...

 

Among the candidate applications:

 

...

- Experimental RasPi clusters.  I have a background in parallelism and concurrency, this is going to be fun.

...

Yes! A cluster of RPi's :-)

I want to do this as well - would be great for experimenting with parallelizable algorithms, distributed <whatever> systems... How often do you get the chance to test the stuff you write on a cluster of machines? Not often in my world - with the RPi I can easily set up 10 machines and start experimenting :-)

Cool stuff!

Jesper Juhl wrote:

 

Morgaine Dinova wrote:

 

...

 

Among the candidate applications:

 

...

- Experimental RasPi clusters.  I have a background in parallelism and concurrency, this is going to be fun.

...

Yes! A cluster of RPi's :-)

I want to do this as well - would be great for experimenting with parallelizable algorithms, distributed <whatever> systems... How often do you get the chance to test the stuff you write on a cluster of machines? Not often in my world - with the RPi I can easily set up 10 machines and start experimenting :-)

Cool stuff!

About the ARM11 architecture: it is not designed for large calculations even an intel atom will probably outperform the processing power of 10 Raspi's see http://infocenter.arm.com/help/topic/com.arm.doc.ddi0333h/index.html..

ed, your answer to Jesper and myself seems to imply that you equate clusters with High Performance Computing.  Well some clusters are used for HPC, certainly, but that's not all a cluster can do.  A cluster is a general purpose computing architecture that can run any program that doesn't require shared state between processors, since the nodes of a normal cluster share no memory.

There are countless non-HPC applications that can run on clusters.  For a start, any application that runs on a single machine but wants a backup node in case of failure, or which wants multiple replicas for load sharing, can be run on a cluster.  In these cases there is still a 1:1:1 mapping between a user, an application and a processor, so there is no concurrency in the actual application.  Clusters work fine in such a role, and are simple to configure.

More ambitious non-HPC cluster applications are those which can harness the parallelism available in a cluster to speed up the application (as in HPC), or to give the application high availability through redundancy, or most recently, to run applications in fully concurrent but non-SMP languages.

The latter are becoming quite popular nowadays.  Erlang has been in use for decades for writing applications that demand extremely high availability, and the same Erlang program will run happily on single core, multicore, or on clusters, making use of as many cores as are available.  Many other concurrent languages have joined Erlang recently as a result of the "multicore revolution", including Scala, Clojure, Haskell, F#, Go, and lots of others, and most of them run as happily on clusters as on SMP multicore systems.

There's a lot of fun to be had with clusters outside of HPC, and Raspberry Pi is certainly a viable candidate for it.

ed, your answer to Jesper and myself seems to imply that you equate clusters with High Performance Computing.  Well some clusters are used for HPC, certainly, but that's not all a cluster can do.  A cluster is a general purpose computing architecture that can run any program that doesn't require shared state between processors, since the nodes of a normal cluster share no memory.

There are countless non-HPC applications that can run on clusters.  For a start, any application that runs on a single machine but wants a backup node in case of failure, or which wants multiple replicas for load sharing, can be run on a cluster.  In these cases there is still a 1:1:1 mapping between a user, an application and a processor, so there is no concurrency in the actual application.  Clusters work fine in such a role, and are simple to configure.

More ambitious non-HPC cluster applications are those which can harness the parallelism available in a cluster to speed up the application (as in HPC), or to give the application high availability through redundancy, or most recently, to run applications in fully concurrent but non-SMP languages.

The latter are becoming quite popular nowadays.  Erlang has been in use for decades for writing applications that demand extremely high availability, and the same Erlang program will run happily on single core, multicore, or on clusters, making use of as many cores as are available.  Many other concurrent languages have joined Erlang recently as a result of the "multicore revolution", including Scala, Clojure, Haskell, F#, Go, and lots of others, and most of them run as happily on clusters as on SMP multicore systems.

There's a lot of fun to be had with clusters outside of HPC, and Raspberry Pi is certainly a viable candidate for it.