Scale Scientific Applications Research Articles

Design and Performance Characterization of RADICAL-Pilot on Leadership-Class Platforms

Many extreme scale scientific applications have workloads comprised of a large number of individual high-performance tasks. The Pilot abstraction decouples workload specification, resource management, and task execution via job placeholders and late-binding. As such, suitable implementations of the Pilot abstraction can support the collective execution of large number of tasks on supercomputers. We introduce RADICAL-Pilot (RP) as a portable, modular and extensible pilot-enabled runtime system. We describe RP's design, architecture and implementation. We characterize its performance and show its ability to scalably execute workloads comprised of tens of thousands heterogeneous tasks on DOE and NSF leadership-class HPC platforms. Specifically, we investigate RP's weak/strong scaling with CPU/GPU, single/multi core, (non)MPI tasks and Python functions when using most of ORNL Summit and TACC Frontera. RADICAL-Pilot can be used stand-alone, as well as the runtime for third-party workflow systems.

“MYCO SYNTHESIS OF SILVER NANOPARTICLES USING GLIOCLADIUM ROSEUM (CLONOSTACHYS ROSEA (LINK) SCHROERS, SAMUELS) AND ITS ANTIMICROBIAL EFFICACY AGAINST SELECTED PATHOGENS”

Objective: Antimicrobial efficacy of silver nanoparticles from Gliocladium roseum, culture filtrate (C. F.) and mycelial mat extract (M. E.) against selected pathogens.
 Methods: Culture filtrate (C. F.) and Mycelial mat extract (M. E.) of Gliocladium roseum were subjected to 10 Mm silver nitrate solution for the synthesis of silver nanoparticles. Formed silver nanoparticles were evaluated via UV-vis spectroscopy and the structural elucidation was done by FT-IR and TEM. Antimicrobial efficacy was tested against bacterial (Salmonella typhi and Klebsiella pneumonia) and fungal (Cladosporium cladosporioides and Alternaria alternata) pathogens. Different nanoparticle concentrations-50, 100, 150 and 200 µl were checked via disc diffusion method.
 Results: Gliocladium roseum (C. F. and M. E.) on interaction with silver nitrate solution effectively reduced metallic silver exhibiting a colour change from yellow to dark brown within 24 h due to the formation of silver nanoparticles. The UV-vis spectrum of C. F. and M. E. showed maximum absorption peaks at 350-400 nm and 400-450 nm respectively and FT-IR and TEM showed strong N-H bonding and spherical shaped silver nanoparticles with the size of 11-19 nm (C. F.) and 25-38 nm (M. E.). Antimicrobial analysis resulted in efficient inhibitory activity against Salmonella typhi, Klebsiella pneumonia and also showed moderate inhibitory activity against Alternaria alternata and Cladosporium cladopsorioides.
 Conclusion: The synthesis of silver nanoparticles from fungus Gliocladium roseum is simple, cheap, safe and eco-friendly thus emphasising on large scale scientific application.

A Data Placement Strategy Based on Crow Search Algorithm in Cloud Computing

Background: In cloud computing era, large scale scientific applications process large amount of data in the data centers. Placing of this data onto a data center is a critical issue performed as part of workflow management system and aims to find the best data center to place the data. It has a direct impact on performance, cost and execution time of workflows. Methods: In this paper, a novel data placement strategy is proposed based on Crow Search Algorithm (CSA) that dynamically distributes the data sets to appropriate data centers during runtime stage of workflow. Results: The results obtained b y CSA based run time data placement algorithm are evaluated with the results of other algorithms. Simulation results acknowledge that using CSA based data placement algorithm provides appropriate results in comparison to the other algorithms.

Data-Oriented Scheduling with Dynamic-Clustering Fault-Tolerant Technique for Scientific Workflows in Clouds

Cloud computing is one of the most prominent parallel and distributed computing paradigm. It is used for providing solution to a huge number of scientific and business applications. Large scale scientific applications which are structured as scientific workflows are evaluated through cloud computing. Scientific workflows are data-intensive applications, as a single scientific workflow may consist of hundred thousands of tasks. Task failures, deadline constraints, budget constraints and improper management of tasks can also instigate inconvenience. Therefore, provision of fault-tolerant techniques with data-oriented scheduling is an important approach for execution of scientific workflows in Cloud computing. Accordingly, we have presented enhanced data-oriented scheduling with Dynamic-clustering fault-tolerant technique (EDS-DC) for execution of scientific workflows in cloud computing. We have presented data-oriented scheduling as a proposed scheduling technique. We have also equipped EDS-DC with Dynamic-clustering fault-tolerant technique. To know the effectiveness of EDS-DC, we compared its results with three well-known enhanced heuristic scheduling policies referred to as: (a) MCT-DC, (b) Max-min-DC, and (c) Min-min-DC. We considered scientific workflow of CyberShake as a case study, because it contains most of the characteristics of scientific workflows such as integration, disintegration, parallelism, and pipelining. The results show that EDS-DC reduced make-span of 10.9% as compared to MCT-DC, 13.7% as compared to Max-min-DC, and 6.4% as compared to Min-min-DC scheduling policies. Similarly, EDS-DC reduced the cost of 4% as compared to MCT-DC, 5.6% as compared to Max-min-DC, and 1.5% as compared to Min-min-DC scheduling policies. These results in respect of make-span and cost are highly significant for EDS-DC as compared with above referred three scheduling policies. The SLA is not violated for EDS-DC in respect of time and cost constraints, while it is violated number of times for MCT-DC, Max-min-DC, and Min-min-DC scheduling techniques.

Genetic-based Multi-criteria Workflow Scheduling with Dynamic Resource Provisioning in Hybrid Large Scale Distributed Systems

Nowadays, scientific progress in multiple disciplines gave rise to conducting large scale scientific experiments and applications known as High Performance Computing (HPC). These HPC applications are commonly structured as workflows of heavy tasks with large data size and intricate dependencies which are typically performed in large-scale distributed systems (LSDS) such as clusters, Grids and recently Cloud infrastructures. In fact, workflow scheduling in distributed systems, especially in Clouds, is proved to be an NP hard problem. Our main target in this paper is to design a workflow scheduling approach based on the Non-dominated Sorting Genetic Algorithm version 2 (NSGA-II) in hybrid distributed systems by optimizing the Makespan and cost. In this work, we also studied the improvement of the Makespan-Cost trade-off with the scalability concept in the Cloud with our designed approach. For that, we proposed different scenarios dealing with the provisioning strategy of processing nodes alongside an existing resource pool. Conducted experiments show the advantage of Cloud infrastructures against other distributed systems and allow investigating the different factors in correlation with resources provisioning process.

Ориентированное на данные планирование с применением отказоустойчивого метода динамической кластеризации для поддержки потоков научных работ в облаках

Cloud computing is one of the most prominent parallel and distributed computing paradigm. It is used for providing solution to a huge number of scientific and business applications. Large scale scientific applications which are structured as scientific workflows are evaluated through cloud computing. Scientific workflows are data-intensive applications, as a single scientific workflow may consist of hundred thousands of tasks. Task failures, deadline constraints, budget constraints and improper management of tasks can also instigate inconvenience. Therefore, provision of fault-tolerant techniques with data-oriented scheduling is an important approach for execution of scientific workflows in Cloud computing. Accordingly, we have presented enhanced data-oriented scheduling with Dynamic-clustering fault-tolerant technique (EDS-DC) for execution of scientific workflows in Cloud computing. We have presented data-oriented scheduling as a proposed scheduling technique. We have also equipped EDS-DC with Dynamic-clustering fault-tolerant technique. To know the effectiveness of EDS-DC, we compared its results with three well-known enhanced heuristic scheduling policies referred to as: (a) MCT-DC, (b) Max-min-DC, and (c) Min-min-DC. We considered scientific workflow of CyberShake as a case study, because it contains most of the characteristics of scientific workflows such as integration, disintegration, parallelism, and pipelining. The results show that EDS-DC reduced make-span of 10.9% as compared to MCT-DC, 13.7% as compared to Max-min-DC, and 6.4% as compared to Min-min-DC scheduling policies. Similarly, EDS-DC reduced the cost of 4% as compared to MCT-DC, 5.6% as compared to Max-min-DC, and 1.5% as compared to Min-min-DC scheduling policies. These results in respect of make-span and cost are highly significant for EDS-DC as compared with above referred three scheduling policies. The SLA is not violated for EDS-DC in respect of time and cost constraints, while it is violated number of times for MCT-DC, Max-min-DC, and Min-min-DC scheduling techniques.

A hybrid multi-objective Particle Swarm Optimization for scientific workflow scheduling

Now-a-days, Cloud computing is a technology which eludes provision cost while providing scalability and elasticity to accessible resources on a pay-per-use basis. To satisfy the increasing demand of the computing power to execute large scale scientific workflow applications, workflow scheduling is the main challenging issue in Infrastructure-as-a-Service (IaaS) clouds. As workflow scheduling belongs to NP-complete problem, so, meta-heuristic approaches are more preferred option. Users often specified deadline and budget constraint for scheduling these workflow applications over cloud resources. But these constraints are in conflict with each other, i.e., the cheaper resources are slow as compared to the expensive resources. Most of the existing studies try to optimize only one of the objectives, i.e., either time minimization or cost minimization under user specified Quality of Service (QoS) constraints. But due to the complexity of workflows and dynamic nature of cloud, a trade-off solution is required to make a balance between execution time and processing cost. To address these issues, this paper presents a non-dominance sort based Hybrid Particle Swarm Optimization (HPSO) algorithm to handle the workflow scheduling problem with multiple conflicting objective functions on IaaS clouds. The proposed algorithm is a hybrid of our previously proposed Budget and Deadline constrained Heterogeneous Earliest Finish Time (BDHEFT) algorithm and multi-objective PSO. The HPSO heuristic tries to optimize two conflicting objectives, namely, makespan and cost under the deadline and budget constraints. Along with these two conflicting objectives, energy consumed of created workflow schedule is also minimized. The proposed algorithm gives a set of Pareto Optimal solutions from which the user can choose the best solution. The performance of proposed heuristic is compared with state-of-art multi-objective meta-heuristics like NSGA-II, MOPSO, and ε-FDPSO. The simulation analysis substantiates that the solutions obtained with proposed heuristic deliver better convergence and uniform spacing among the solutions as compared to others. Hence it is applicable to solve a wide class of multi-objective optimization problems for scheduling scientific workflows over IaaS clouds.

Cost-Time Efficient Scheduling Plan for Executing Workflows in the Cloud

The emergence of Cloud Computing as a model of service provisioning in distributed systems instigated researchers to explore its pros and cons on executing different large scale scientific applications, i.e., Workflows. One of the most challenging problems in clouds is to execute workflows while minimizing the execution time as well as cost incurred by using a set of heterogeneous resources over the cloud simultaneously. In this paper, we present, Budget and Deadline Constrained Heuristic based upon Heterogeneous Earliest Finish Time (HEFT) to schedule workflow tasks over the available cloud resources. The proposed heuristic presents a beneficial trade-off between execution time and execution cost under given constraints. The proposed heuristic is evaluated for different synthetic workflow applications by a simulation process and comparison is done with state-of-art algorithm i.e. BHEFT. The simulation results show that our proposed scheduling heuristic can significantly decrease the execution cost while producing makespan as good as the best known scheduling heuristic under the same deadline and budget constraints.

Implementation of a Multi-threaded Framework for Large- scale Scientific Applications

The CMS experiment has recently completed the development of a multi-threaded capable application framework. In this paper, we will discuss the design, implementation and application of this framework to production applications in CMS. For the 2015 LHC run, this functionality is particularly critical for both our online and offline production applications, which depend on faster turn-around times and a reduced memory footprint relative to before. These applications are complex codes, each including a large number of physics-driven algorithms. While the framework is capable of running a mix of thread-safe and "legacy" modules, algorithms running in our production applications need to be thread-safe for optimal use of this multi-threaded framework at a large scale. Towards this end, we discuss the types of changes, which were necessary for our algorithms to achieve good performance of our multithreaded applications in a full-scale application. Finally performance numbers for what has been achieved for the 2015 run are presented.

Intelligent failure prediction models for scientific workflows

The ever-growing demand and heterogeneity of Cloud Computing is garnering popularity with scientific communities to utilize the services of Cloud for executing large scale scientific applications in the form of set of tasks known as Workflows. As scientific workflows stipulate a process or computation to be executed in the form of data flow and task dependencies that allow users to simply articulate multi-step computational and complex tasks. Hence, proactive fault tolerance is required for the execution of scientific workflows. To reduce the failure effect of workflow tasks on the Cloud resources during execution, task failures can be intelligently predicted by proactively analyzing the data of multiple scientific workflows using the state of the art of machine learning approaches for failure prediction. Therefore, this paper makes an effort to focus on the research problem of designing an intelligent task failure prediction models for facilitating proactive fault tolerance by predicting task failures for Scientific Workflow applications. Firstly, failure prediction models have been implemented through machine learning approaches using evaluated performance metrics and also demonstrates the maximum prediction accuracy for Naive Bayes. Then, the proposed failure models have also been validated using Pegasus and Amazon EC2 by comparing actual task failures with predicted task failures.

A data‐centric framework for debugging highly parallel applications

SummaryContemporary parallel debuggers allow users to control more than one processing thread while supporting the same examination and visualisation operations of that of sequential debuggers. This approach restricts the use of parallel debuggers when it comes to large scale scientific applications run across hundreds of thousands compute cores. First, manually observing the runtime data to detect error becomes impractical because the data is too big. Second, performing expensive but useful debugging operations becomes infeasible as the computational codes become more complex, involving larger data structures, and as the machines become larger. This study explores the idea of a data‐centric debugging approach, which could be used to make parallel debuggers more powerful. It discusses the use of ad hoc debug‐time assertions that allow a user to reason about the state of a parallel computation. These assertions support the verification and validation of program state at runtime as a whole rather than focusing on that of only a single process state. Furthermore, the debugger's performance can be improved by exploiting the underlying parallel platform because the available compute cores can execute parallel debugging functions, while a program is idling at a breakpoint. We demonstrate the system with several case studies and evaluate the performance of the tool on a 20 000 cores Cray XE6. Copyright © 2013 John Wiley & Sons, Ltd.

Chunked extendible dense arrays for scientific data storage

Several meetings of the Extremely Large Databases Community for large scale scientific applications advocate the use of multidimensional arrays as the appropriate model for representing scientific databases. Scientific databases gradually grow to massive sizes of the order of terabytes and petabytes. As such, the storage of such databases require efficient dynamic storage schemes where the array is allowed to arbitrarily extend the bounds of the dimensions. Conventional multidimensional array representations cannot extend or shrink their bounds without relocating elements of the data-set. In general, extendibility of the bounds of the dimensions, is limited to only one dimension. This paper presents a technique for storing dense multidimensional arrays by chunks such that the array can be extended along any dimension without compromising the access time for an element. This is done with a computed access mapping function, that maps the k-dimensional index onto a linear index of the storage locations. This concept forms the basis for the implementation of an array file of any number of dimensions, where the bounds of the array dimension can be extended arbitrarily. Such a feature currently exists in the Hierarchical Data Format version 5 (HDF5). However, extending the bound of a dimension in the HDF5 array file can be unusually expensive in time. Such extensions, in our storage scheme for dense array files, can still be performed while still accessing elements of the array at orders of magnitude faster than in HDF5 or conventional arrays-files. We also present theoretical and experimental analysis of our scheme with respect to access time and storage overhead. Such mapping scheme can be readily integrated into existing PGAS models for parallel processing in a cluster networked computing environment.

Parallel multi-dimensional range query processing with R-trees on GPU

The general purpose computing on graphics processing unit (GP-GPU) has emerged as a new cost effective parallel computing paradigm in high performance computing research that enables large amount of data to be processed in parallel. Large scale scientific data intensive applications have been playing an important role in modern high performance computing research. A common access pattern into such scientific data analysis applications is multi-dimensional range query, but not much research has been conducted on multi-dimensional range query on the GPU. Inherently multi-dimensional indexing trees such as R-Trees are not well suited for GPU environment because of its irregular tree traversal. Traversing irregular tree search path makes it hard to maximize the utilization of massively parallel architectures. In this paper, we propose a novel MPTS (Massively Parallel Three-phase Scanning) R-tree traversal algorithm for multi-dimensional range query, that converts recursive access to tree nodes into sequential access. Our extensive experimental study shows that MPTS R-tree traversal algorithm on NVIDIA Tesla M2090 GPU consistently outperforms traditional recursive R-trees search algorithm on Intel Xeon E5506 processors.

Cloud computing infrastructure for biological echo-systems

The Biological applications such as Gene and Protein analysis integrate and analyze biological data for the research in many bioinformatics and other bio related fields. Such applications are used under many large scale scientific applications and help in computing, integrating data, execute the analysis, automate the process by using information retrieved by different tasks and computational procedures to assist the scientists in scientific discovery and data distribution. Grid based and/or web based scientific workflow tools are used for bioinformatics related complex research to make scientists’ and researchers’ work easier. On average, scientists spend about 80% of their time assembling data to prepare for analysis. This is due largely in part to the fact that many of these resources required for data processing must be gathered from an external source. The best of these resources, however, are scattered across the globe. They are hosted at universities, institutes, and laboratories throughout the world. To bring all of these resources together by hiding system, network, and application level heterogeneity issues are challenging.

Using GPU's to accelerate stencil-based computation kernels for the development of large scale scientific applications on heterogeneous systems

We present CaCUDA - a GPGPU kernel abstraction and a parallel programming framework for developing highly efficient large scale scientific applications using stencil computations on hybrid CPU/GPU architectures. CaCUDA is built upon the Cactus computational toolkit, an open source problem solving environment designed for scientists and engineers. Due to the flexibility and extensibility of the Cactus toolkit, the addition of a GPGPU programming framework required no changes to the Cactus infrastructure, guaranteeing that existing features and modules will continue to work without modification. CaCUDA was tested and benchmarked using a 3D CFD code based on a finite difference discretization of Navier-Stokes equations.

Advance Bandwidth Scheduling Algorithms in Dedicated Networks

An increasing number of high performance research network testbeds and production networks have the capability of provisioning dedicated channels for high speed data transfer in support of large scale scientific applications. Each dedicated channel in these networks typically consists of one or more physical links that are shared by multiple applications in both time and bandwidth through in advance reservation. The design of efficient bandwidth scheduling algorithms is critical to maximizing the utilization of dedicated network resources and meeting diverse end-to-end transport performance requirements. The topology of a dedicated network is represented by a graph, where each link maintains a list of residual bandwidths specified as segmented constant functions of time. Given a graph with an aggregated time-bandwidth table combining the reservation information on all links, designated source, and destination vertices, and a specified data size, we formulate and investigate five bandwidth scheduling problems to minimize the data transfer end time under different path and bandwidth constraints: (i) variable path with variable bandwidth (VPVB), which computes the widest (highest bandwidth) path in each time slot; (ii) fixed path with variable bandwidth (FPVB), which computes a fixed path with varying bandwidths across different time slots; (iii) variable path with fixed bandwidth (VPFB), which computes one path in each time slot with the same bandwidth; (iv) fixed path with fixed bandwidth (FPFB), which computes a fixed path with a constant bandwidth; and (v) multiple fixed paths with fixed bandwidths (MFPFB), which computes multiple concurrent fixed paths with constant bandwidths. Among these problems, VPVB represents the most flexible scenario where the network resources are fully utilized and the minimal transfer end time is achieved, while FPFB imposes the most stringent transport conditions by fixing both path and bandwidth. We design an optimal algorithm for each of these scheduling problems. VPVB is solved using an extension of the classical Dijkstra's algorithm, and FPVB is solved using Maximum Permutation Algorithm, which tries all possible permutations of throughputs at different time slots to obtain the minimal transfer end time. The algorithm for VPFB, FPFB, and MFPFB are based on Adjacent Time Slots Search, which finds the minimum transfer end time by examining the possibility of data transfer in any number of contiguous time slots. All these algorithms are of polynomial or pseudo polynomial time complexity with respect to the network size and the total number of time slots in a bandwidth reservation table.

NEKTAR, SPICE and Vortonics: using federated grids for large scale scientific applications

In response to a joint call from US's NSF and UK's EPSRC for applications that aim to utilize the combined computational resources of the US and UK, three computational science groups from UCL, Tufts and Brown Universities teamed up with a middleware team from NIU/Argonne to meet the challenge. Although the groups had three distinct codes and aims, the projects had the underlying common feature that they were comprised of large-scale distributed applications which required high-end networking and advanced middleware in order to be effectively deployed. For example, cross-site runs were found to be a very effective strategy to overcome the limitations of a single resource. The seamless federation of a grid-of-grids remains difficult. Even if interoperability at the middleware and software stack levels were to exist, it would not guarantee that the federated grids can be utilized for large scale distributed applications. There are important additional requirements for example, compatible and consistent usage policy, automated advanced reservations and most important of all co-scheduling. This paper outlines the scientific motivation and describes why distributed resources are critical for all three projects. It documents the challenges encountered in using a grid-of-grids and some of the solutions devised in response.

Library for architecture-independent development of structured grid applications

P++ is an innovative user environment that simplifies the development of efficient, parallel programs for large scale scientific applications. It permits portability across the widest variety of computer architectures. The target machines are distributed memory computers with different types of node architectures (scalar, vector, or superscalar), but the requirements of shared memory computers are also addressed. Unfortunately, robust, architecture-independent and efficient software environments have been notable to date primarily by their absence. Such a simplifying environment for the development of software is sorely needed to take advantage of the current and future developments in advanced computational hardware. The P++ environment does this using a standard language, C++, with absolutely no modification of the compiler. For parallel communication it employs existing widely portable communications libraries. Such an environment allows existing C++ language compilers to be used to develop software in the preferred serial environment, and such software to be efficiently run, unchanged, in all target environments.The explicit goal of P++ is the support of advanced computational methods for the solution of large-scale computational problems. For efficiency reasons and also for simplification, the use of P++ is restricted to the large class of structured grid methods for partial differential equations. Applications using P++ can selectively exploit the added degree of freedom presented by parallel processing by use of an optimization module within the array language interface. A collection of defaults ensures deterministic behavior. In this way, complicated optimization issues of parallel execution may be easily introduced by the setting of switches within the user interface, which are then interpreted at runtime. Through the use of Virtual Shared Memory, restricted to array variables (Virtual Shared Grids), issues such as partitioning become matters of optimization, and not criteria for correct execution. Due to the restriction and optimization for structured grids, we expect the same parallel performance as for codes based on the traditionally used explicit Message Passing model. To speed the development of the P++ environment, we use an efficient, commercially available array language library, M++, developed in C++; the M++ array interface is also used as the P++ array interface.Recognizing that the acceptance of advanced parallel computers depends on their use by scientists, and not parallel researchers, it is hoped that P++ will simplify access to these parallel machines. While significantly simplifying our own research on parallel self adaptive local refinement methods, it is hoped that P++ will more generally simplify the introduction of structured grid methods for large scale scientific computing onto advanced parallel machines.

An integrated development and run-time environment for distributed memory parallel computers

After presenting a short overview of existing portable software tools supporting the implementation of codes for full scale scientific applications, we present the concept of the integrated development and run-time environment SPADE currently under development, which will not only support Monte Carlo simulations of lattice gauge theory, but also application areas like physical chemistry, image analysis and neural networks. The key features of our system include portability to a range of heterogeneous networks of processors and in particular expandability, where the latter also allows to supplement completely new application areas. We conclude by presenting the current status and the time schedule for the development of SPADE.

A performance interval for the micro-mainframe relationship for scientific and mathematical programming applications

The performances as measured in virtual CPU time of an IBM PC/XT and an 8 MHz compatible system are compared with a 3081D mainframe computer. The results are used to link the performance of a microcomputer to the family of IBM mainframes for scientific and mathematical programming applications. Through the use of the t-test, a 99% confidence interval is established to determine the expected performance of a microcomputer relative to the family of IBM mainframes. The test problems used in this study consisted of 14 zero-one goal programming problems found in the literature. The results of this study provide an improved basis for the micro-mainframe decision for larger scale scientific and mathematical programming applications.