Research On Resouse Load Balance Strategy Based On Swift

With the rapid development of Internet technology, Internet of Things technology and widespread of smart mobile devices, data generate all the time, now we have entered the era of big data. The rapid growth of massive data bring a lot of challenge to data storage technology, traditional storage technology can not deal with it well, so could storage technology appears. As one kind of cloud storage system, Swift can meet the requirements of massive data stor age, but should be optimized in performance and resource utilization. Load balance technology plays an important role in improve system performance and resource utilization, so it has important practical significance to study how to improve performance and resource utilization by load balance.Firstly, by study of Swift’s read strategy, we know that in order to ensure the reliability of data storage, each data will be stored in multiple copies. When downloads data, it will randomly select one storage node which has the copy of the data. This read strategy does not consider the load status of storage node, so there may be a large number of requests on some storage nodes while the other storage nodes are relatively free. In order to improve system’s resource load balance and performance, this paper puts forward a load balance read strategy based on storages resource usage. Proxy node selects the least loaded storage node to read according to the received feedback of storage node’s I/O utilization. Experiment results show that, compared with Swift’s random read strategy, the read strategy proposed by this paper enables the storage nodes I/O utilization and CPU utilization achieve better load balance status.Secondly, by study Swift’s proxy node, we know proxy node is the entrance for outside to access Swift cloud storage system. The proxy node will become single points of failure and performance bottlenecks when a large number of user requests. But Swift has no proxy node load balance strategy. This paper puts for ward a strategy that achieve Swift proxy node load balance based on Linux Virtual Server. Since the load balance scheduling algorithm plays an important role in load balance, this paper studies load balance scheduling algorithm further, puts forward a dynamic feedback load balance algorithm based on weighted least connection scheduling. Load balance node adjusts the weights of proxy nodes dynamically according to proxy nodes’ CPU utilization, memory utilization and network utilization in order to distribute load more reasonably. Experiments show that, compared with no proxy node load balance, achieve proxy nodes resource load balance based on Linux Virtual Server can improve system throughput and bandwidth, reduce response time. Compared with weighted least connections scheduling, dynamic feedback load balance algorithm based on weighted least connection scheduling can further improve system throughput and bandwidth, reduce response time.