面向云应用的存储缓存子系统

发布时间：2018-03-15 02:21

本文选题：缓存　切入点：云应用　出处：《计算机科学》2014年04期 　论文类型：期刊论文

【摘要】：性能是云应用的重要指标,云应用与现有应用在数据存储与管理方式和访问模式等方面存在较大差异,这使得传统的缓存管理算法难以适应云应用的要求。针对云应用的特性,设计了基于影响因子的缓存策略,亦即将元数据和数据缓存分开管理,使用影响因子综合管理影响缓存被再次访问的几率的多种因素,区别创建、打开、读取和修改等操作对缓存再次访问的几率的影响;设计了缓存关联管理策略,即利用元数据和数据之间的关联提高缓存管理的性能;设计了缓存主动调度策略,即通过主动淘汰较低影响因子的缓存项和动态调整元数据与数据缓存的大小来提高缓存子系统的适应能力和性能。最后实现了原型系统,并使用Filebench和Postmark进行了测试和分析,验证了面向云应用缓存子系统原型能提高1%~120%的I/O性能以及2%~87%的操作处理速度。
[Abstract]:Performance is an important index of cloud applications. There are great differences between cloud applications and existing applications in data storage and management methods and access patterns, which makes it difficult for traditional cache management algorithms to adapt to the requirements of cloud applications. The cache strategy based on the influence factor is designed, that is, the metadata and data cache are managed separately, and the influence factors are used to synthetically manage the factors that affect the probability of the cache being accessed again. The influence of reading and modifying operations on the probability of cache re-access; the design of cache association management strategy, which uses the association between metadata and data to improve the performance of cache management; and the design of cache active scheduling strategy. That is, to improve the adaptability and performance of the cache subsystem by actively eliminating the cache items with lower impact factors and dynamically adjusting the size of metadata and data cache. Finally, the prototype system is implemented and tested and analyzed by using Filebench and Postmark. It is verified that the prototype of the cloud-oriented cache subsystem can improve the I / O performance by 120% and the operating speed of 2% by 87%.
【作者单位】：江苏大学计算机科学与通信工程学院;
【基金】：高等学校博士学科点专项科研基金(20093227110005) 广东省自然科学基金(S2011010006118) 江苏省高校自然科学基金(09KJB520001) 浙江省自然科学基金(LY13F020012)资助
【分类号】：TP333

【相似文献】