ErlCee++ (and starting remopte erlang nodes)

Joe Armstrong <>
Tue Jun 8 14:35:59 CEST 2004


> Joe has often remarked that many of the nice properties of Erlang
> processes are shared by Unix processes, chiefly isolation. He flatly
> dismisses actually using OS ones just because he's an efficiency nut :-)

No - I'm an isolation nut - I'm well known for my efficient code :-)

> 
> I'd really like to see an Erlang-like runtime system built on Unix
> processes. How far towards the Erlang programming style could it be
> pushed? Wouldn't it be fun to be able to write "Erlang" processes in
> any language and mix 'em all together?

Well one way to do this might be to use PVM.

<aside>

  The  other day  I awoke  with the  thought "It's  all  about message
passing".

  Q: How do we write distributed stuff?
  A: Send asynchronous messages - you'll never know if they arrive
     if you *must* know - send a reply

  Note no shared anything :-) - easy to understand etc

  How can I do this in a  big network with firewalls etc. Can A send a
message to B if A and B are *both* behind firewalls etc. (yes send via
a proxy C - but both A and B must have agreed to communicate via some
proxy)

  I   asked  my  old   friend  Mr.   Google  about   "message  passing
infrastructures" and he told me about PVM ...

</aside>

Here are some quotes from http://www.csm.ornl.gov/pvm/intro.html

  " PVM  is a message  passing system that  enables a network  of Unix
computers  to  be  used   as  a  single  distributed  memory  parallel
computer. This network is referred to as the virtual machine....

  In  order to write  a parallel  program, tasks  must be  executed on
different processors....

  ... Here is an example of a typical call to pvm_spawn().

  numt=pvm_spawn("my_task", NULL, PvmTaskDefault, 0, n_task, tids)

  This spawns n_task copies of  the program "my_task" on the computers
that PVM  chooses. The actual number  of tasks started  is returned to
numt. The  task id  of each task  that is  spawned is returned  in the
integer  array tids.   Every PVM  process  has a  unique integer  that
identifies it--this is its task id....

  ...  After  the data have been  packed into the  sending buffer, the
message is  ready to  be sent.   This is accomplished  with a  call to
pvm_send().

  info=pvm_send(tid, msgtag)

  will send  the data in  the sending buffer  to the process  with the
task id of tid....

  The receiving task makes a  call to pvm_recv() to receive a message.

  bufid=pvm_recv(tid, msgtag)

  will wait for a message from task tid with tag msgtag to arrive, and
will receive it when it does.

  Does any of this sound familiar?

  Note that pvm_spawn spawns ANY  process (ie not an Erlang process) -
so this could be used to remotely start an erlang process processes on
another node (ie it solve the problem raised in another thread).

  PVM also has some nice introspection facilities (Goggle XPVM)

  I haven't thought  about it that much but it appears  that a nice way
of interfacing  Erlang with "foreign"  applications would be  to teach
Erlang to speak  pvm - either by writing a linked  in driver to allow
Erlang to  send and  receive pvm messages  or by implementing  the PVM
protocol in Erlang  itself.

 If you Google pvmbook.pdf you can find a very readably book about PVM.

 Cheers

  /Joe

  [PS] If PVM is  not the answer then what is? -  All I want is simply
messag passing  from A to B.  TCP  is too low-level and  Erlang is too
specific.

  I just want a simple C library to send and receive messages - handle
failures  connect  firewalled machines  via  proxies  and have  simple
(configurable) naming and security policies.
  
   


> Side notes:
> 
>   Linux 2.6 does one thousand context-switches per second, spends
>   absolutely bugger-all CPU to do it, and AFAIK the scheduler is O(1).
> 
>   Memory is cheap.
> 
> Cheers,
> Luke
> 





More information about the erlang-questions mailing list