Rajeev Wankar

Smart Replica Selection for Data Grids Using Rough Set Approximations (RSDG)

The best replica selection problem is one of the important aspects of data management strategy of data grid infrastructure. Recently, rough set theory has emerged as a powerful tool for problems that require making optimal choice amongst a large enumerated set of options. In this paper, we propose a new replica selection strategy using a grey-based rough set approach. Here first the rough set theory is used to nominate a number of replicas, (alternatives of ideal replicas) by lower approximation of rough set theory. Next, linguistic variables are used to represent the attributes values of the resources (files) in rough set decision table to get a precise selection cause, some attribute values like security and availability need to be decided by linguistic variables (grey numbers) since the replica mangers’ judgments on attribute often cannot be estimated by the exact numerical values (integer values). The best replica site is decided by grey relational analysis based on a grey number. Our results show an improved performance, compared to the previous work in this area.

Topology Aware Task Stealing for On-chip NUMA Multi-core Processors☆

Abstract “The On Chip NUMA Architectures (OCNA) introduce a new challenge namely memory-latency to the scheduling methods. The language run-times and libraries try to explore the processing power of these multiple cores by mapping the user-created tasks on to these cores by using suitable scheduling algorithms with load balancing support to improve throughput. The popular load balancing techniques used are work-sharing and work-stealing and many run-time systems such as Cilk, TBB and wool implement task stealing algorithm to schedule the tasks on to the cores by multiplexing the program generated tasks on to the native worker threads supported by the operating system. But the task stealing strategy applied in present run-time systems assumes the sharing the last level cache (LLC) and common shared bus among all cores on Chip Multi Processor. It tries to optimize the utilization without considering the presence of multiple On Die DRAM controllers and their topological arrangements. Current task stealing technique also suffers from problem of randomly choosing the victim worker queue. In this paper we address these issues and propose a solution for these problems by suggesting few optimizations. Our proposed task stealing strategy dynamically analyzes the topology of the underlying hardware connections and models the group of cores and connections as a logical topology tree. This logical tree is translated into multiple worker pools called stealing domains. By restricting the task stealing within these domains, this strategy is implemented and shows an average of 1.24 times better performance on NAS Parallel Benchmark programs compared to popular runtimes Cilk and OpenMP.

Replica Selection in Data Grids Using Preconditioning of Decision Attributes by K-means Clustering (K-RSDG)

This paper extends the applicability of the Rough Set Replica Selection Strategy in Data Grids, RSDG, proposed previously for such situations where the history of replica sites is unavailable. Grey based Rough Set Theory is applied using replicas information only as an input data. The Decision attributes are derived applying Grey based K-means clustering algorithm upon the input data. Each cluster label of K clusters represents a class of Decision attribute in the Decision Table of the rough set. Comparing to the previous work the synthetic data experimentation shows the improvement in the overall performance.

Design of a Structured Fine-Grained Access Control Mechanism for Authorizing Grid Resources

The heterogeneous and very dynamic nature of a grid environment demands a very scalable authorization system. This brings out the need for a fast fine-grained access control (FGAC) mechanism for grid resources to accomplish the required scalability. This paper is proposing a new method of storing and manipulating the resources security policy in a way that enables authorization systems to avoid redundancy in security policy checking and to more rapidly specify which group of resources the authenticated user is authorized to access. Comparative study is made in a simulated environment. The proposed method can be integrated in the present authorization systems for better performance.

A two phased service oriented Broker for replica selection in data grids

Replica selection is one of the fundamental problems in Data Grids environment. This works concern is designing a Two phasedService Oriented Broker (2SOB) for replica selection. It is focused on investigating, selecting, modifying, and experimenting with some non-conventional approaches to be applied on the relevant selection techniques. The motivation of this work is to introduce a novel Service OrientedBroker for Replica Selection in Data Grid. The main characteristics of this 2SOB are: Scalability, Reliability, Availability, Efficiency and Ease of deployment. 2SOB consists of two phases; the first is a Coarse-grainphase, basically used for sifting replica sites that have low latency (uncongested network links) and distinguishing them from other replicas having a high latency (congested network links). Procedurally, this has been done using the association rules concept of the Data Mining approach. The second is a Fine-grainphase,used for extracting the replicas admissible for user requirements through applying Modified Minimum Cost and Delay Policy (MMCD). Both phases have accordingly been designed, simulated, coded, and then validated using real data from EU Data Grid.The first phase has thereby been applied on the real network data of CERN (February 2011). Experimentations compared with some other contemporary selection methods of different Brokers showed appreciable results. Using this proposed Broker it is possible to achieve an enhancement in the speed of executing Data Grid jobs through reducing the transfer time.

Performance Prediction on Multi-core Processors

Contention for shared resources between co-running programs on a multi-core processor causes degradation of performance. Previous research work made efforts to characterize and classify memory behavior of programs to predict their performance. Such knowledge may be used to design workloads for performance studies on multi-core processors. It could also be utilized to formulate policies that alleviate contention for shared resources at a system level. In this work we apply machine learning techniques to predict the performance on multi-core processors. The main contribution of the study is enumeration of solo-run program attributes, which can be used to predict paired-run performance. The paired run involves the contention for shared resources between co-running programs. We observed that machine learning techniques could be utilized to predict the paired run performance with reasonable accuracy.

Rough set clustering approach to replica selection in data grids (RSCDG)

In data grids, the fast and proper replica selection decision leads to better resource utilization due to reduction in latencies to access the best replicas and speed up the execution of the data grid jobs. In this paper, we propose a new strategy that improves replica selection in data grids with the help of the reduct concept of the Rough Set Theory (RST). Using Quickreduct algorithm the unsupervised clustering is changed into supervised reducts. Then, Rule algorithm is used for obtaining optimum rules to derive usage patterns from the data grid information system. The experiments are carried out using Rough Set Exploration System (RSES) tool.

A methodology for high availability of data for business continuity planning / disaster recovery in a grid using replication in a distributed database

High availability systems ensure nonstop service delivery despite natural calamities or any other disruptions, whatsoever. In a grid, high availability can be achieved by offering services of a failed node from any other live node, if redundant storage of data and software of the failed node is made available on that live node. During replication of data in a grid for high availability, the techniques used in distributed databases can be exploited. Data replication can be achieved better by efficient data distribution across the various nodes in the grid. In this paper, a methodology for high availability of data for business continuity planning/disaster recovery (BCP/DR) in a grid environment is proposed, using some of the features of distributed query processing. With this approach OGSA-DQP (open grid services architecture - distributed query processing) standard can be extended to OGSAHA for high availability.

Rough set based quality of service design for service provisioning in clouds

Quality of Service (QoS) is a broad term used to describe the overall experience a user or application will receive over a network. A rough set based approach is used to design a modified Cloud-QoS Management Strategy (MC-QoSMS). MC-QoSMS is a component of cloud broker that is used to allocate resources based on Service Level Agreement between users and providers for Infrastructure as a Service (IaaS) provisioning of cloud. Concept of reduct from rough set theory is used to allocate the best service provider to the clouds user with minimum searching time. The performance of the proposed system has been analyzed in terms of number of requests. It is reported that the system outperformed random algorithm by 25% and the round robin algorithm by 30% for 100 requests.

Enhancing the hierarchical clustering mechanism of storing resources’ security policies in a grid authorization system

Many existing grid authorization systems adopt an inefficient structure of storing security policies for the available resources, which reduces the scalability and leads to huge repetitions in checking security rules. One of the efficient mechanisms that handles these repetitions and increases the scalability is the Hierarchical Clustering Mechanism (HCM) [1]. HCM outperforms the Brute Force Approach as well as the Primitive Clustering Mechanism (PCM). This paper enhances HCM to accommodate the dynamism of the grid and the same is demonstrated using new algorithms.