Hi Prakhyat, Assuming the OS scheduler is fair, if all threads are busy then all the computational resources will be distributed roughly equally.
> Consider 3 dispatchers with 8 threads each on a dual core machine. > > I am not sure what you mean by each on a dual core machine. Each of the actor systems? Anyway, taking a more interesting example: There are: - 8 cores - 2 dispatchers - one dispatcher having 12 threads hosting 100 actors - second dispatcher having 4 threads hosting 1000 actors If one core is 100% then we have 800% available on the machine. There are 16 threads in total, each will get roughly 50% (800/16). This means that the first dispatcher will get roughly 12*50% = 600% while the second will get 50%*4 = 200%. Of course this is only true if both dispatchers have active actors, for example if the second one is idle, then the first one will get all the 800%. Also, the first dispatcher hosts 100 actors, so you will get roughly 6% for each actor (if all actors are busy) and for the second dispatcher you will get 0.2% for each actor (if all actors are busy). Of course if not every actor is busy (very common) then the active ones will get a much larger share. -Endre > -Prakhyat M M > > > On Thursday, 28 August 2014 15:17:42 UTC+5:30, Akka Team wrote: > >> Hi Prakhyat, >> >> >> >> On Thu, Aug 28, 2014 at 11:24 AM, Prakhyat Mallikarjun < >> [email protected]> wrote: >> >>> Hi Andre, >>> >>> Thanks. >>> >>> For point [1] What do you mean by "It is also a very good practice to >>> use separate dispatchers for various parts of the system." >>> >>> I have not seen any app with separate dispatchers. Kindly share any >>> article which describes this and app over github would be help. >>> >> >> Akka itself is using separate dispatchers. For example, of configured so, >> clustering and remoting can run on their own dispatchers, separately from >> the default dispatcher (so these subsystems do not need to compete with >> user actors for the same set of threads). Also Akka IO uses dedicated >> dispathers to handle IO (again, configurable). >> >> This is a somewhat old article about dispatchers: >> http://letitcrash.com/post/40755146949/tuning-dispatchers-in-akka- >> applications >> >> Also I believe most of the Akka books discuss dispatchers and their >> importance. >> >> -Endre >> >> >>> >>> I have my app using akka persistence/remoting/sharding. >>> >>> -Prakhyat M M >>> >>> On Thursday, 28 August 2014 14:44:13 UTC+5:30, Akka Team wrote: >>> >>>> Hi Prakhyat, >>>> >>>> >>>> >>>> We are creating one actor system for entire app. Correct me if I am >>>>>> wrong, the default dispatcher by default uses a fork-join thread >>>>>> pool with at least 8 threads. >>>>>> >>>>>> We need to support millions of requests from client. How to tune >>>>>> dispatcher of actor system when using akka persistence/sharding/ >>>>>> clustering, >>>>>> >>>>>> 1. What is Recommended Type of Dispatcher >>>>>> >>>>>> Usually the ForkJoin pool is the best option for most of the >>>> workloads, but this is something you need to measure. It is also a very >>>> good practice to use separate dispatchers for various parts of the system. >>>> Then they might be tuned individually. >>>> >>>> >>>>>> 1. How to calculate threads per thread pool >>>>>> >>>>>> This is documented in the configuration file: >>>> >>>> # The parallelism factor is used to determine thread pool size >>>> using the >>>> # following formula: ceil(available processors * factor). >>>> Resulting size >>>> >>>> >>>> >>>> # is then bounded by the parallelism-min and parallelism-max >>>> values. >>>> >>>> >>>> >>>>> >>>>>> 1. Do we need to consider cores? >>>>>> >>>>>> Yes, for example if you have only 8 cores then configuring 100 >>>> threads in the pool makes usually no sense. If you have 8 cores then you >>>> might also want to have at least 8 threads. >>>> >>>> >>>>> >>>>>> 1. Do we need to consider type of OS >>>>>> >>>>>> I don't have any benchmarks of Akka that compares different OSes on >>>> the same machine. Also, this is probably workload dependent. >>>> >>>> >>>>> >>>>>> 1. Will JVM configuration matter >>>>>> >>>>>> Yes of course. Again, it is not easy to give any particular >>>> configuration, these are all use case depedent. >>>> >>>> >>>>> [Note:] Suggest any missed parameter in above list. Also how to >>>>>> specify custom dispatcher configuration in app conf. >>>>>> >>>>>> Basically want to understand criteria's to decide performing >>>>>> dispatcher per actor system. >>>>>> >>>>>> Is there any other configuration required to tune akka >>>>>> persistence/cluster/shard kind of application for good performance. >>>>>> >>>>>> Also recommendations for tuning akka persistence/cluster/shard kind >>>>>> of application would be great help. >>>>>> >>>>> >>>> Apart from very generic advice we cannot really do much more without >>>> detailed information about your application. If you want in-depth expert >>>> advice tailored to your needs you might want to look into our consulting >>>> offerings: http://typesafe.com/how/consulting >>>> >>>> -Endre >>>> >>>> >>>>> >>>>>> -Prakhyat M M >>>>>> >>>>>> -- >>>>> >>>>>>>>>> Read the docs: http://akka.io/docs/ >>>>> >>>>>>>>>> Check the FAQ: http://doc.akka.io/docs/akka/c >>>>> urrent/additional/faq.html >>>>> >>>>>>>>>> Search the archives: https://groups.google.com/grou >>>>> p/akka-user >>>>> --- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "Akka User List" group. >>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>> an email to [email protected]. >>>>> To post to this group, send email to [email protected]. >>>>> >>>>> Visit this group at http://groups.google.com/group/akka-user. >>>>> >>>>> For more options, visit https://groups.google.com/d/optout. >>>>> >>>> >>>> >>>> >>>> -- >>>> Akka Team >>>> Typesafe - The software stack for applications that scale >>>> Blog: letitcrash.com >>>> Twitter: @akkateam >>>> >>> -- >>> >>>>>>>>>> Read the docs: http://akka.io/docs/ >>> >>>>>>>>>> Check the FAQ: http://doc.akka.io/docs/akka/ >>> current/additional/faq.html >>> >>>>>>>>>> Search the archives: https://groups.google.com/ >>> group/akka-user >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "Akka User List" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at http://groups.google.com/group/akka-user. >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> >> >> -- >> Akka Team >> Typesafe - The software stack for applications that scale >> Blog: letitcrash.com >> Twitter: @akkateam >> > -- > >>>>>>>>>> Read the docs: http://akka.io/docs/ > >>>>>>>>>> Check the FAQ: > http://doc.akka.io/docs/akka/current/additional/faq.html > >>>>>>>>>> Search the archives: https://groups.google.com/group/akka-user > --- > You received this message because you are subscribed to the Google Groups > "Akka User List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/akka-user. > For more options, visit https://groups.google.com/d/optout. > -- Akka Team Typesafe - The software stack for applications that scale Blog: letitcrash.com Twitter: @akkateam -- >>>>>>>>>> Read the docs: http://akka.io/docs/ >>>>>>>>>> Check the FAQ: >>>>>>>>>> http://doc.akka.io/docs/akka/current/additional/faq.html >>>>>>>>>> Search the archives: https://groups.google.com/group/akka-user --- You received this message because you are subscribed to the Google Groups "Akka User List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/akka-user. For more options, visit https://groups.google.com/d/optout.
