Virtualization Overhead Research Articles

LGF: A flexible framework for exposing legacy codes as services

Scientific applications, usually written in Fortran or C, need to be adapted to work in modern environments for conducting scientific computations, which are based on web services or components. This work presents a Legacy to Grid framework (LGF) which enables semi-automatic virtualization of legacy codes as grid services. In contrast to existing work, LGF goes beyond a simple adapter-pattern approach; it follows a two-tier design in which the adaptation service layer is loosely coupled with the legacy back end layer. We present the architecture of our framework, its basic operation, and its capabilities: support for different interaction patterns with a legacy code and for computation migration to enable fault-tolerance and dynamic load balancing of legacy codes. We also present a case study of our framework’s usage. We conclude with an evaluation of the impact of the two-layer architecture on the application’s performance and the tradeoff between flexibility and performance by measuring the overhead of virtualization for different granularities of the legacy code.

Implications of virtualization on Grids for high energy physics applications

The simulations used in the field of high energy physics are compute intensive and exhibit a high level of data parallelism. These features make such simulations ideal candidates for Grid computing. We are taking as an example the GEANT4 detector simulation used for physics studies within the ATLAS experiment at CERN. One key issue in Grid computing is that of network and system security, which can potentially inhibit the widespread use of such simulations. Virtualization provides a feasible solution because it allows the creation of virtual compute nodes in both local and remote compute clusters, thus providing an insulating layer which can play an important role in satisfying the security concerns of all parties involved. However, it has performance implications. This study provides quantitative estimates of the virtualization and hyper-threading overhead for GEANT on commodity clusters. Results show that virtualization has less than 15% run time overhead, and that the best run time (with the non-SMP license of ESX VMware) is achieved by using one virtual machine per CPU. We also observe that hyper-threading does not provide an advantage in this application. Finally, the effect of virtualization on run time, throughput, mean response time and utilization is estimated using simulations.

Scale and performance in the Denali isolation kernel

This paper describes the Denali isolation kernel, an operating system architecture that safely multiplexes a large number of untrusted Internet services on shared hardware. Denali's goal is to allow new Internet services to be "pushed" into third party infrastructure, relieving Internet service authors from the burden of acquiring and maintaining physical infrastructure. Our isolation kernel exposes a virtual machine abstraction, but unlike conventional virtual machine monitors, Denali does not attempt to emulate the underlying physical architecture precisely, and instead modifies the virtual architecture to gain scale, performance, and simplicity of implementation. In this paper, we first discuss design principles of isolation kernels, and then we describe the design and implementation of Denali. Following this, we present a detailed evaluation of Denali, demonstrating that the overhead of virtualization is small, that our architectural choices are warranted, and that we can successfully scale to more than 10,000 virtual machines on commodity hardware.