Cloud Storage

In this study we consider large scale, distributed storage systems with a redundancy mechanism. The goal of the project is to investigate how such systems can reduce their power consumption during low-utilization time intervals by operating in a low-power mode. In a low power mode, a subset of the disks or nodes are powered down, yet we ask that each data item remains accessible in the system; this is called full coverage. Our main objective is to incorporate this option into existing systems rather than redesign a system. When doing so, it is crucial that the low power option should not affect the performance or other important characteristics of the system during full-power (normal) operation.

Our research addresses this question for generic distributed storage systems (where the key component under investigation is the placement function of the system) as well as for specific popular system designs in the realm of distributed storage. The observations and techniques are instrumental for a wide spectrum of systems, ranging from storage systems for the enterprise to cloud data services. In the cloud environment where low cost is imperative, the effects of such savings are magnified by the large scale.

Among our observations and results are:

We produce a general framework for obtaining full coverage by using auxiliary disks.  Find a fraction p of nodes that when powered down still leaves a cover of all but a (small) q fraction of the data. The q fraction is then covered by auxiliary disks We analyze the power savings possible in a generic system (both positive and negative results) We demonstrate that with random placement one cannot do better than the generic case

One of the most widely used placement functions, the consistent hashing, readily accommodates a low power mode. See graph for possible power-down set vs. covered data tradeoffs In a hierarchical system it is sufficient to modify the lower level placement in order to achieve optimal power savings