1

Topic: Planning of network sharing

There is  the project it very modular in leaking  it it is launched as one process but in the future  it is required for reliability that units were in  processes at the given stage if it  that it strongly all complicates because  and  aspects often  as it would be possible to consider at a design stage of the one-process version at least partially future sharing on separate processes? How to organize  so that then to minimize changes connected with that that hundreds communications should be ?

2

Re: Planning of network sharing

Hello, Kingofastellarwar, you wrote: K> there is  project K> it very modular K> in leaking  it it is launched as one process K> but in the future  it is required for reliability that units were in  processes K> at the given stage if it  that it strongly all complicates because  and  aspects often  K> as it would be possible to consider at a design stage of the one-process version at least partially future sharing on separate processes? K> how to organize  so that then to minimize changes connected with that that hundreds communications should be ? And does that it that? As variants to unwinding: 1) sharding with retrimming 2)  the server. The useful activity in that case , so it is possible to balance from the client (or a separate proxy-balansera with which the client communicates?). The server (it the server?) In that case simply fulfills certain actions on the data which has come to it. Generally without particulars heavily something to tell...

3

Re: Planning of network sharing

Hello, Kingofastellarwar, you wrote: K> it very modular K> how to organize  so that then to minimize changes connected with that that hundreds communications should be ? For this purpose units should interact with each other through a message exchange. Through the bus or directly. It is possible to look on Akka and ZeroMQ

4

Re: Planning of network sharing

Hello, Kingofastellarwar, you wrote: K> how to organize  so that then to minimize changes connected with that that hundreds communications should be ? Stages: 1) to fix interfaces of units, operation with units to conduct only through interfaces 2) to make calls of interfaces asynchronous on the basis of queue (each unit receives the queue of "messages") 3) to make instead of several queues for units one general, to add in messages the information on the unit-sender and the unit-appointment for dispatching 4) to carry out queue in separate process, to implement protocols of operation with queue normally it makes sense to postpone points 2,3,4 to  removal of units in processes, for queues it is better to take ready decisions, for example redis

5

Re: Planning of network sharing

Hello, Kingofastellarwar, you wrote: K> how to organize  so that then to minimize changes connected with that that hundreds communications should be ? The Keyword "then". To solve which problems now is not present - a direct way to create these problems. Is better do not do anything. When it is required to space apart something on different processes at you there will be an exact understanding as the decision should look and you quickly implement it.

6

Re: Planning of network sharing

Hello, gandjustas, you wrote: G> Hello, Kingofastellarwar, you wrote: K>> how to organize  so that then to minimize changes connected with that that hundreds communications should be ? G> the Keyword "then". G> to solve which problems now is not present - a direct way to create these problems. G> is better do not do anything. When it is required to space apart something on different processes at you there will be an exact understanding as the decision should look and you quickly implement it. I will dare to disagree. However, as generic rule of thumb "then" - descends, but Unforeseen possibilities on clustering and sharding cost dearer later. I.e. if it is authentically known that is expected , it is necessary to do a cluster, ,  etc., will be cheaper if  to be enclosed at once. The contribution can be a miscellaneous, for example: - Even on one DB the server it is possible to dissolve the data under different circuits and not to use FK between different circuits ( each circuit leaves on the server) - to avoid JOIN between tables of different circuits - not to lay down to the logician in a DB - to do stateless services - communication between services - only through the high-level interface (assuming that it in the future is transformed to exterior calls of the same REST or that select) well and from such spirit

7

Re: Planning of network sharing

Hello, Kingofastellarwar, you wrote: K> how to organize architecture Architecture of corporate program applications. Martin Fowler

8

Re: Planning of network sharing

Hello, velkin, you wrote: V> Hello, Kingofastellarwar, you wrote: K>> how to organize architecture V> Architecture of corporate program applications. Martin Fowler on  there interesting is not present anything if what to recommend for a subject that on microservices, there questions of network sharing are considered explicitly