一种基于块级的存储性能优化方法

Method for Storage Performance Optimization at Block Level

性能是存储管理系统中的一项重要内容,提高存储性能可以有效地降低系统的总体拥有成本(TCO)。数据迁移是性能优化的重要的组成部分,确定一个有效的模型,以决定迁移的对象和迁移的位置非常重要。提出一个定量的评价模型,与仅关心IO特性的热点检测系统不同,该模型考虑多方面的影响因素,评估磁盘中每个数据块的重要性,以此作为数据分类的依据。与其他模型最大的不同在于它是基于块级的,不考虑文件系统和上层应用程序,因此具有较好的性能。实验表明,该模型可以用于热点检测,指导数据迁移;有助于提高存储系统中存储设备的利用率。

Performance is a key factor in storage management systems. To improve storage performance could reduce the TCO dramatically. As one of the most important elements in performance optimization, data migration needs a more reasonable model to decide what and where to migrate. A measurable method for evaluating the value of every data block in a disk was proposed. Comparing with the hotspots detecting system that focuses on IO characters, more factors were taken into consideration in this model. To establish the model at block level overlooking the file system and application, is the most obviously difference to other model, so it can help a storage system to achieve high performance. It can be concluded from the experiment that this method is helpful for the hotspots detection, to guide the data migration. In addition, it could also help to enhance the utilization of storage devices in the storage system.