[kepler-dev] best practices for minimizing process start-up latency?
Edward Lee
eal at eecs.berkeley.edu
Fri Jan 10 10:42:58 PST 2014
I agree that aspect-oriented modeling is the right way to go...
The prior checkpointing work is described here:
http://chess.eecs.berkeley.edu/pubs/389.html
and
http://ptolemy.eecs.berkeley.edu/publications/papers/06/incrementalCheckpointing_WSC06/index.htm
Edward
On 1/10/14 10:08 AM, Christopher Brooks wrote:
> Ah! I see, I misunderstood.
>
> Yes, it should be possible to use Aspect Oriented programming to do
> this. Page 370 of the book has some information about aspect oriented
> programming.
>
> I believe you would need to add an actor that would handle the
> recording and then annotate the ports that are to use that actor.
>
> There is this larger issue of checkpointing and rollback that could be
> useful for long running models. The idea is that if we could
> checkpoint a model, then if there was a problem with execution, we
> could resume from the checkpoint. If the actors have no state, then
> knowing what tokens are on the ports might be sufficient. However, if
> the actors have state, then more work would probably be necessary.
> There was some work done with modifying actor code to allow
> backtracking, though not much has happened with that code for some time.
>
> _Christopher
>
> On 1/10/14 9:44 AM, Rich Morin wrote:
>> On Jan 10, 2014, at 09:20, Christopher Brooks wrote:
>>> My responses are below.
>> Thanks!
>>
>>>> The following questions have to do with the inter-actor "piping"
>>>> and best practices for command debugging and optimization.
>>>>
>>>> Q: Can I add recording and/or archiving as attributes to a pipe?
>>>>
>>>> So, for example, could I tell Kepler to turn on recording for
>>>> particular pipes, without needing to explicitly add an actor?
>>>> (This could be used for "tracing" key paths in an app.)
>>> I don't know of an actor that has access to pipes. ...
>> Perhaps I should have said "plumbing". I'm referring to the usual
>> communication channels used by Kepler/Ptolemy actors, not to Unix
>> named pipes, etc.
>>
>> There was a talk at the PII meeting about hanging attributes on
>> connections. I was wondering if something like this could be used
>> to enable archiving or analysis of the transmitted data, etc.
>>
>> -r
>>
>> --
>> http://www.cfcl.com/rdm Rich Morin rdm at cfcl.com
>> http://www.cfcl.com/rdm/resume San Bruno, CA, USA +1 650-873-7841
>>
>> Software system design, development, and documentation
>>
>>
>
>
More information about the Kepler-dev
mailing list