Arutyun Avetisyan

Open Cirrus: A Global Cloud Computing Testbed

Open Cirrus is a cloud computing testbed that, unlike existing alternatives, federates distributed data centers. It aims to spur innovation in systems and applications research and catalyze development of an open source service stack for the cloud.

Two level job-scheduling strategies for a computational grid

We address parallel jobs scheduling problem for computational GRID systems. We concentrate on two-level hierarchy scheduling: at the first level broker allocates computational jobs to parallel computers. At the second level each computer generates schedules of the parallel jobs assigned to it by its own local scheduler. Selection, allocation strategies, and efficiency of proposed hierarchical scheduling algorithms are discussed.

Virtualizing HPC applications using modern hypervisors

In this paper we explore the prospects of virtualization technologies being applied to high performance computing tasks. We use an extensive set of HPC benchmarks to evaluate virtualization overhead, including HPC Challenge, NAS Parallel Benchmarks and SPEC MPI2007. We assess KVM and Palacios hypervisors and, with proper tuning of hypervisor, we reduce the performance degradation from 10-60% to 1-5% in many cases with processor cores count up to 240. At the same time, a few tests provide overhead ranging from 20% to 45% even with our enhancements. We describe the techniques necessary to achieve sufficient performance. These include host OS tuning to decrease noise level, using nested paging with large pages for efficient guest memory allocation, and proper NUMA architecture emulation when running virtual machines on NUMA hosts. Comparing KVM/QEMU and Palacios hypervisors, we conclude that in general the results with proper tuning are similar, with KVM providing more stable and predictable results while Palacios being much better on fine-grained tests at a large scale, but showing abnormal performance degradation on a few tests.

Improving Properties of a Parallel Program in ParJava Environment

ParJava integrated environment supporting development and maintenance of data parallel Java-programs is discussed. When a parallel program is developed it is necessary to assure not only its correctness, but also its efficiency and scalability. For this purpose it is useful to know some dynamic properties of the program. This information may help to modify the program in order to improve its parallel features. ParJava provides a collection of such tools. Symbolic execution allows to estimate expected execution time and limits of scalability of Java SPMD-program.

Static analyzer Svace for finding defects in a source program code

This paper describes Svace, a tool for static program analysis developed at the Institute for Systems Programming, Russian Academy of Sciences. This tool allows one to find defects and potential vulnerabilities in the source program code written in C/C++ languages. The main features of the tool are simplicity of use, wide variety of supported types of warnings, scalability up to programs of millions of code lines, and acceptable quality of analysis (30–80% of true positive warnings).

Estimation of dynamical characteristics of a parallel program on a model

This paper considers a model of a parallel program that can efficiently be interpreted on an instrumental computer, providing a means for a sufficiently accurate prediction of the actual time needed for execution of the parallel program on a given parallel computational system. The model is designed for parallel programs with explicit message passing written in Java with calls to the MPI library and is a part of the ParJava environment. The model is obtained by transforming the program control tree, which, for Java programs, can be constructed by modifying the abstract syntax tree. The communication functions are simulated on the basis of the LogGP model, which makes it possible to take into account specific features of the distributed computational system.

Possibilities of Optimal Execution of Parallel Programs Containing Simple and Iterated Loops on Heterogeneous Parallel Computational Systems with Distributed Memory

The problem of load balancing when executing parallel programs on computational systems with distributed memory is currently of great interest. The most general statement of this problem is that for one parallel loop: execution of a heterogeneous loop on a heterogeneous computational system. When stated in this way, the problem is NP-complete even in the case of two nodes, and no acceptable heuristics for solving it are found. Since the development of heuristics is a rather complicated task, we decided to examine the problem by elementary methods in order to refine (and, possibly, simplify) the original problem statement. The results of our studies are discussed in this paper. Estimates of efficiency of parallel loop execution as functions of the number of nodes of homogeneous and heterogeneous parallel computational systems are obtained. These estimates show that the use of heterogeneous parallel systems reduces the efficiency even in the case when their communication subsystems are scaleable (see the definition in Section 4). The use of local networks (heterogeneous parallel computational systems with nonscaleable communication subsystems) for parallel computations with heavy data exchange is not advantageous and is possible only for a small number of nodes (about five). An algorithm of optimal distribution of data between the nodes of a homogeneous or heterogeneous computational system is suggested. Results of numerical experiments substantiate the conclusions obtained.

Scalable and accurate detection of code clones

A detailed description of a method for detection of code clones is described. This method is based on the semantic analysis of programs and on new algorithms that make it scalable without affecting its accuracy. The proposed method involves two phases. In the first phase, the program dependence graph (PDG) is constructed while the program is compiled. LLVM is used as the compilation infrastructure. In the second phase, similar subgraphs of maximum size that represent code clones are detected. Before starting the search for similar subgraphs, the PDG is divided into subgraphs that will be considered as potential clones of each other. To ensure scalability of the search for similar subgraphs, the composition of algorithms is used. The first algorithm checks that a pair of graphs cannot have similar subgraphs of the desired size; this is done in a linear amount of time. If this algorithm fails, another (approximate) algorithm is executed to find similar subgraphs of maximum size. After similar subgraphs have been found, the program code is additionally checked for the position of the code lines corresponding to the detected clone candidates. Tests showed that the developed tool is more accurate than similar tools, such as MOSS, CCFinder, and CloneDR. Results obtained for the projects Linux-2.6, Firefox Mozilla, LLVM/Clang, and OpenSSL are presented.

Productivity prediction of MPI programs based on models

A model of parallel program that can be effectively interpreted on the development computer guaranteeing the possibility of a sufficiently precise prediction of real run time for a simulated parallel program at the prescribed computer system is studied. The model is worked out for parallel programs with explicit message passing written in the Java language with MPI library access and is included into the composition of ParJava environment. The model is obtained by transforming the program control tree that can be constructed for Java programs by modifying the abstract syntax tree. To model communication functions, the model LogGP is used which allows taking into consideration the specific character of the communication network of the distributed computer system.

Extension of Java Environment by Facilities Supporting Development of SPMD Java-Programs

ParJava is extension of Java environment by facilities to support developing of effective scalable portable object-oriented parallel SPMD-programs for homogenous and heterogeneous computing systems with distributed memory. ParJava model is determined by four Java-interfaces, which support the notion of computing space as JavaVM network. ParJava allows to perform parallel programs designed for homogenous system on heterogeneous ones without lost of scalability. These facilities provide base supporting design of high-level object models of parallel programming.