The Design And Implementation Of Performance Benchmark Test System For Cloud Object Storage

With the development of Internet, data generated every day ha ve shown explosivegrowth. In response to numerous types of data types and the trend of rapid growth, moreand more people began turning to object storage service system. Currently there havebeen many object storage services with similar feature and interface. And because ofdifferent mechanisms for object storage services and the lack of a common workloadsimulation tool for object storage applications to fetch performances of the storageservice system (such as throughput, response time, success ratio for accessing, CPUusage, memory usage, etc.), it’s hard for users of storage services to do horizontal andvertical comparison to propose a solution which suits themselves best.The vacancy of workload simulation tool in the field of object storage application ishoped to be filled by the design and implementation of performance benchmark testSystem for cloud object storage which can be applied to a variety of object storagesystems. For different storage systems, users just need realize that storage system adapterinterface to dynamically load it into the benchmark platform service. To customize anddistribute the test load, the user can easily use XML file to interact with the Controller ofplatform. After the platform Agent being scheduled by Controller to run test workload,real-time display of instantaneous performance metrics will be stored in the WEB front.Once all workloads are finished, the implementation process will archive all the variousstorage performance metrics on the local disk for user to analyze and compare them.Users can not only use these performance metrics to tune the same storage systemservices continuously, but also can use the same physical environment to test andcompare the performance of different storage systems services.Currently benchmark platform has been put into use, it can support OpenStack Swift,Amazon S3and other object storage systems. This system provides users with an effective benchmark for tuning and performance optimization, contributing to thedevelopment of cloud object storage performance tuning research work. And it can alsobe used for cloud object storage architecture design, deployment solutions and so on.