Research On Performance Optimization For Online Social Networks

With the rapid development of online social networks, online social networks are on the type of increasingly diverse. The number of users is also on the rise. Every user of online social networks can independent publish personalized data, such as blogs, pictures, videos, and tweets, and so on. It makes the data of online social networks become larger and larger. Online social networks present the following characteristics:the great amount of data and high speed of growth, and data diversity, dynamic and difference of users. These characteristics of online social networks, trigger a series of new system performance bottlenecks, which greatly restrict the development of online social networks. How to solve the performance bottleneck is crucial.Information retrieval of online social networks is namely high privacy text retrieval problem; Concurrent access problem of online social networks is the concurrent access problem caused by a large number of users of microblogging platform; Information storage problem of online social networks is the storage problem with valid data under the mobile social networks. These problems have to be faced in the development process of online social networks. It is the key difficult problem and must be solved.Information retrieval of online social networks is different from the traditional web search. The data of online social networks has a high degree of privacy and use the Key-Value of storage model. Information retrieval based on text content is very difficult. In view of this characteristic, this paper proposes an efficient text search optimization method based on the user summary index. A lightweight friend summary index table structure is proposed in this paper. The text contents of two hop friends of users are mapped into a summary index table. It is efficient to solve data privacy protection issues in online social networks, while maintaining the retrieval efficiency. It designs two levels of sorting algorithm, according to the user summary index table. It can filter out the unnecessary friends access, avoid invalid access caused by exhaustive queries, and greatly reduce the communication cost between servers. Then, via a kind of approximate TF×IDF text sorting algorithm, in the case of maintaining high retrieval accuracy, it can improve the retrieval efficiency and reduce the retrieval latency. Experimental results show that the text search optimization method based on the user summary index can reduce94.1%of the network communication overhead caused by the exhaustive queries, decrease82.4%of latency, and maintain high retrieval accuracy.Online social networks, especially the microblog platform, are playing more and more the roles of the news media platform. Users gain information of important events from the microblog platform, such as earthquake relief, large sports events, etc. These significant events or emergencies cause the users who focus on the events increasing explode in a short period of time, which resulting in a large number of concurrent access to the servers. It is easy to cause congestion, and even lead to collapse of the system platform. This paper proposes a concurrent access optimization method based on a p2p structure, to solve load balance and scalability issues caused by large-scale concurrent access of microblog platform when the emergency occurs. By using the SoMed systems structure, according to the differences of user behavior, the users of microblog platform can be divided into media users and social users. According to this classification, it designs a system structure of two levels of DHT. All members of the media users constitute the first level of DHT, the members of the social users constitute the second layer of DHT, and the fan clique is the basic unit. The second layer of DHT is constituted with the fan clique centralized with a media user. Each fan clique is made up of a media user and his fans. Through the way of the peer assistance in the same clique, it achieves the goal of reducing concurrent access problems. Experimental results show that the optimization method based on the two levels of DHT can greatly reduce the bandwidth costs caused by the concurrent access problem when an emergency occurs in microblog system, and reduce the network latency with high system stability and availability.With the rapid growth of friend numbers of online social networks, the amount of information gained from the newsfeeds is becoming larger and larger. When the storage capacity is limited, especially online social networks gradually moving to the mobile terminals, the information filtering is particularly important. Studies show that the friend relationship strength of users actually changes over time. Most relationships of friends never contact after a month. The dynamic information generated by this kind of "strange friends" is not interested at all. This paper presents an information storage optimization method based on the interaction relationship strength of friends, to solve information filtering problem under the environment of mobile social networks. Put forward a kind of adjustable counting bloom filter to record user interaction relationship between friends. Information generated by the friends with little interaction will be filtered out. Experimental results show that the information storage optimization method based on the interaction between friends can effectively filter out the invalid information, greatly reduce the storage cost with high filtration accuracy.To sum up, through the studies of retrieval efficiency, concurrency control, and information filtering performance optimization in online social network, it can improve the server performance of online social networking platform, improve the quality of user services, and have great research significance for the development of online social network.