| Traffic checkpoints data has characteristics of multi- source, high speed and large volume,etc. At present the traffic bayonet system is still using the traditional relational database. While dealing with these traffic bayonet data, this database has exposed the problems of poor expansibility, low storage capacity and low query efficiency. Faced with these problems, in this thesis a HBase-based storage and query system was designed and implemented for traffic bayonet data.The system used a distributed architecture, front camera sensor based on the Http protocol send data to Flume. It used Flume-ng to collect, aggregate, and move large amounts of traffic bayonet data in order to meet the real-time demand. Then, it moved different kinds of data to Kafka as distributed message queue in data-partition. Storm is a distributed computing system which got real-time data from the message queue and complete storage process. At last, storm wrote to HBase clusters with Storm. Using HBase as Sql on the Phoenix-client layer, we can achieve the HBase database query. Within that context, we should ensure the system with high reliability, high availability to realize the system which could read and write quickly.In the implementation of the test, we used the RD650 ThinkServer to build a multi-node distributed server platform. The experimental results demonstrate that, compared with storage system based on Oracle, this system which was designed has 3 times increment in storage performance. Compared with the original HBase, it also has 3 times increment. Furthermore, it supports multi-conditional joint query.Thus, the above scheme can improve currently traffic bayonet system whose performance is poor in reading and writing due to the large amount of data. Moreover, it can be predicted that with the rapid development of the information technology, we will foresee the realization of the wisdom of the traffic in the near future. Deteriorating urban traffic conditions can be effectively improved. |
PDF Full Text Request