Performance analysis of vertically partitioned data in clouds through a client-based in-memory key-value store cache

Abstract

Data security and protection in Cloud Computing are still major challenges. Although Cloud Computing offers a promising technological foundation, data have to be stored externally in order to take the full advantages of public clouds. These challenges lead to our distribution approach that vertically distributes data among various cloud providers. As every provider only gets a small chunk of the data, the chunks are useless without the others. Unfortunately, the actual performance is disillusioning and the access times of the distributed data are indisputable. Thus, this lousy performance is now in the focus of this work. The basic idea is the introduction of a cache that stores the already joined tuples in its memory. Thus, not always the different cloud storages have to be queried or manipulated, but only the faster caches. Finally, we present the implementation and evaluation of a client-based In-Memory cache in this work.