CUDSwap: Tolerating Memory Exhaustion Failures in Cloud Computing

Cloud computing is now being used by a wide variety of users, ranging from expert programmers and system administrators to scientists and laymen. Cloud providers are taking full advantage of all their resources as much as they can. Memory is the most expensive resource in terms of oversubscription and this has resulted in high price to the end user. Furthermore, performing swapping in Virtual Machines (VM) is expensive, so the cloud providers usually do not offer any swapping space for their systems. As a consequence, when a VM runs out of memory, user processes are killed. This scenario in cloud environment is especially critical, since the user loses all of his/her execution time and, by extension, the money invested in this computation. For cloud users such as life scientists with varying memory requirements, this is a critical problem. This paper presents CUD Swap, a kernel extension module designed to detect memory exhaustion in cloud instances, add more swap space, and thus prevent process failures. CUD Swap has been implemented in Linux kernel and has been evaluated over a variety of workloads as well as real-world life science applications. The paper describes CUD Swap design and implementation details, and reports performance details from the evaluation.