Design Of Distributed System With High Availability And Scalability

The number of many online games are typical distributed system,The availability, scalability of system and intermediate task parallelism is difficult subject.Because of the need to support thousands to millions of online game player,and at the same time to deal with all kinds of complicated game logic in order to improve the experience of game player,Then,it could increase the cohesion of the game players.This framework requires that the system has strong scalability, high availability, scalability and development.Because when game is firstly launched,it is impossible to expect the number of game players,All the online games will face problems of scalability.The system need to increase the processing capacity according to the growth of the number of users At the late time of the game,users lose interest and reduce the processing capacity.Although almost games will do maintenance at a specific time,But 7*24 is still on the system requirements,High availability system is still an important quality attribute.The current trend of the chip architecture from increasing clock speed into a nuclear,The designing goal of the new chip is to do multiple things simultaneously,In order to adapt to the change of system chips,it need to achieve the overall performance through concurrent growth.The consistency problem of distributed system is also a big problem for most developers,The best way is to use architecture solve it The subject of this thesis is to design a distributed architecture,The programming model is reactive,Distributed system architecture is also a new type of MMORPG,The goal is to hide the distributed computing as possible to the actual developers, the consistency problem in distributed systems,And still guaranteeing the high availability and scalability of the system.Which makes The end developer can focus on business logic rather than the details of the underlying technology. The achieved research results are as follows:(1) Optimized network communication module, changed the way of traditional network library in single IO threads, realized the multi nuclear reactor model,that is multi IO thread with the thread pool model. For application layer cache is optimized to reduce system calls and using a temporary stack space,so as to improve the system of the network throughput and reduce delay.(2) In order to improve the system availability,two cluster named zookeeper and Redis are introduced, zookeeper solved the distributed system of naming,distributed lock and load balance, coordination and consistency of redis as hotspot data in memory database loading system to ease pressure from traditional disk database, such improvements increased system throughput and reduce delay,because the redis is cluster deployment in the system, but also ensure the system's usability and performance.(3) The task service mentioned in the framework is typical worked structure of master/slave, the task can be in different mobile node, in order to achieve load balance, due to the slave can be dynamically added, using the master to monitor the slave node and realize simple load balancing strategies, so as to ensure the system's scalability.(4) Data services provide data service for the entire system, belonging to the schema for distributed data,which is logically centralized entity, on multiple machines clustering partitioned data, through the control of multi machine on different data partitions, you can achieve load balancing and scalability.