Hi Alexey. Good questions! Answers: a) short term: me (small number of processes). Baby steps. Long term: there needs to be an awareness in each process of at least a subset of the others and something that cues rebirth. Rebirth to an initial configuration would be the duty of another process; b) which data? c) this is part of my learning. In the physical world how is the data represented? When I fake the system it is the equivalent of all processes broadcasting 'to all' or all information being posted to a common 'blackboard'. I thought maybe I could do the same, say using CloudDB (which I'm familiar with) or an equivalent. But it just seemed to me that this would make the DB a bottleneck as it handles information pushed to it and pulled from it. But maybe that's okay? d) see a) e) # of processes is task dependent and self-regulating as the system receives feedback on its performance. Questions of load ... I'm still learning so I've no idea. I'm still very ignorant of real world of the hardware considerations for distributed real-world processing.
You're correct, I can fake it (and have been to a certain degree). The problem is that faking it leads to a predictable, brittle system. Also, a faked system has no asynchrony and has complete information. However, as you may have guessed, the functionality I'm seeking is to yield a dynamic, slightly unpredictable, but potentially adaptive system. So far, I've found that faking the system is hard and part of the reason I've looked to a real construct; probably I'll just be trading one set of challenges for a different set, but I'll be learning something in the process so it's all good :-) > > There are questions like > > - who will keep an eye on monitoring and restarting died agents. > - where to store data? > - what to do if data sent to died agent? Should it be recovered and > restarted? > - how to restart and restore state of died agents? > - how much agents start on one machine? how to measure, distribute and > balance load? How to add new machine or fix broken? > ... > > And all this becomes even more complex if there are requirements for > transactions or atomicity of operations. > > I'd suggest to not to built such system and instead fake it with simpler > approaches. I.e. message queues, databases, stateless workers. > -- -- Job Board: http://jobs.nodejs.org/ Posting guidelines: https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines You received this message because you are subscribed to the Google Groups "nodejs" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/nodejs?hl=en?hl=en --- You received this message because you are subscribed to the Google Groups "nodejs" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
