Evaluating HDFS I/O Performance on Virtualized Systems

Size: px
Start display at page:

Download "Evaluating HDFS I/O Performance on Virtualized Systems"

Transcription

1 Evaluating HDFS I/O Performance on Virtualized Systems Xin Tang University of Wisconsin-Madison Department of Computer Sciences Abstract Hadoop as a Service (HaaS) has received increasing attentions due to its flexibility and low cost to do big data analysis. This project studies the I/O performance of the Hadoop Distributed File System (HDFS) in both native and virtualized systems by implementing a new tool using Chen and Patterson s workload model to benchmark the throughput of HDFS. The benchmark program consists of two main modules: worker and manager that read in four parameters and handle process requests. The entire test is divided in to two experiments, with the second one as the optimized version of the first one. The results in general match the expectation for the virtualized and native systems and they have also invoked new possibility for future optimization. 1. Introduction Nowadays, in the time of booming technological innovations, people demand larger hard drives, larger memory, etc. Things are better if larger. As small data gets outcompeted by big data, Hadoop, featuring big data analysis at a low cost, is brought to spotlight. Currently, an increasing amount of companies have moved their IT infrastructures to the virtualized cloud to lower management cost. When such movement comes together with the demand for big data, a new market is born -- Hadoop as a Service (HaaS) [1]. There are already several companies that provide Hadoop service in the cloud in this market, such as Amazon EMR (Elastic Map Reduce). The HaaS market is rapidly growing even though it is relatively new. It is predicted that the market capital will grow from 131 million dollars in 2012 to 1.9 billion dollars in 2016 [1]. Such huge growing potential indicates the significance and necessity of research on the performance of Hadoop in the cloud. As a big data analysis tool, Hadoop s key component is file system, which deals with storage and transfer of large data files. The performance of file inputs and outputs of the file system will greatly affect the entire Hadoop performance. Therefore, a tool that evaluates IO performance of the Hadoop system would play an important role in improving Hadoop s performance in the cloud. The default and also the most popular benchmark tool in the HDFS is Test DFSIO, which provides basic functionality of benchmarking. However this tool is not comprehensive enough to fully support the optimization of Hadoop s performance in the cloud. To handle the optimization problem, I have implemented a benchmark based on Chen and Patterson s workload model for I/O system [2], which is more suitable in targeting Hadoop optimization in this project. I have tested my implementation in HDFS on both native and virtualized systems and obtained some preliminary results. The rest of this report is organized as follows. In section 2, I provide the related works of this project. Section 3 and 4 introduce my benchmark workload model and implementation. Section 5 and 6 discuss the experiments that I have run. I end this paper with a conclusion of my project and discussion of future work in Section Related Works 2.1 Hadoop and HDFS This section introduces Hadoop and HDFS, which are the objects of my I/O benchmark, to readers that are unfamiliar with these techniques. Hadoop: is an open source framework that implements the MapReduce programming model [3]. It supports data-intensive distributed applications on large clusters of commodity hardware. Since it offers a low cost solution to big data analysis, Hadoop is getting increasingly

2 popular. For example, prominent users of hadoop include Yahoo!, Facebook, and Amazon. Hadoop Distributed File System (HDFS): is a distributed, scalable, and portable file system for the Hadoop framework [4]. HDFS stores large files in multiple of 64 MB and access them sequentially. Since Hadoop is a well-known technique, I only briefly introduce the features that affect the design of my benchmark here. Readers interested in learning more about Hadoop and HDFS are encouraged to read the online documentation [4]. 2.2 TestDFSIO TestDFSIO is the default I/O test tool for HDFS. It allows users to select file size and buffer size and computes the throughput and IO rate of the current Hadoop File System setup. Users can use this as a basic tool to understand the performance of their hadoop clusters. However, this tool also has a few limitations. First, its benchmark model does not consider concurrency. This model assumes that there is only one job doing read or write in the Hadoop system while in real situation, multiple jobs can issue IO requests simultaneously. This model obviously fails to capture this characteristic. Also this test has to run with the MapReduce framework. It cannot be used to benchmark HDFS only. 3. Workload Model The core of any benchmark used in performance evaluation is the workload model. At the suggestion of the smart, handsome and excellent program chair of CADAVER 13, I use the canonical I/O system workload model proposed by Chen and Patterson to provide a good synthetic evaluation of HDFS [2]. This workload model characterizes five parameters that lead to the firstorder performance effects in I/O systems: 1) uniquebytes: the number of unique data byte read and written in a workload. 2) sizemean: the average size of an IO request. 3) readfrac: the fraction of reads. 4) processnum: the number of processes simultaneously issuing IO requests. 5) seqfrac: the fraction of requests that follow the prior request sequentially. This model is comprehensive and easy to implement. Unlike TestDFSIO, which only considers the file and buffer sizes, the parameters of this model also reflect the special locality, temporal locality and concurrency of an I/O system. Since this is a parameterized synthetic benchmark model, the implementation is easy. I can test the effects of a parameter by simply keeping the other ones unchanged. As a final note in this section, I would like to point out that HDFS is designed to process large amount of data sequentially. Hence, although I use this model as the core of my benchmark, the last parameter, seqfrac, is ignored in my implementation. 4. Implementation This section describes my implementation that tailor Chen and Patterson s workload model to run in the Hadoop framework. The benchmark program consists of two main modules: worker and manager. Manager: defines the configurations of the IO tests and hires workers to do File IO in HDFS. It sets the default values of all four parameters, selects the parameters and their ranges to test, and determines the number of runs for each test. During each run, it starts a number of workers in Hadoop. This number is equal to the processnum in the current run. The manager passes the values of uniquebytes, sizemean and readfrac to the workers and waits. After all the workers complete their tasks, it reads their processing time of File IO and computes the average as the File IO time of the current run. Worker: does file read and write in HDFS and records the total processing time of its IO requests. After getting the parameters from the manager, the worker clears the memory cache and starts to read data from or write data to two separated files in HDFS. It issues IO requests repeatedly until the total bytes being read and written is equal to or greater than uniquebytes. To simulate the IO requests in a real system, I use a pseudo random number generator to randomize the order of reading and writing data. This generator returns a number that is chosen pseudo-randomly with uniform distribution between 0 and 1. The worker

3 reads data if the returned number is smaller than or equal to readfrac and writes data otherwise. than in the native because the virtual system was running on top of another OS and it had less memory resource than the native system. 5.2 Results Figure 1. The interactions between the Manager and Workers. 5. First Experiment This section discusses the setup and results of my first experiment of running my benchmark tool in HDFS. 5.1 Setup Due to limited resources, I conducted this experiment on a single laptop with Intel Core 2 (2.4GHz) processor and 4GB 1067 MHz DDR3 memory. The native OS was Mac OS 10.6, the virtualization platform was VirtualBox and the virtualized OS was Ubuntu Since the computer had a limited disk size, I used relatively small values in this experiment. The default values of the parameters are 64 MB uniquebytes, 1 KB sizemean, 1 processnum and 0 readfrac. I tested how each parameter would affect the overall throughput by changing its value multiples times while keeping the other three with the default values. The testing range for each parameter is the following: unique- Bytes (64MB 2GB), sizemean (1KB 2MB), processnum (1 5), and readfrac (0 1). I expected that an increase in uniquebytes size or processnum would decrease the throughput while an increase in sizemean and readfrac would increase the throughput. Also I expected the throughput change caused by these parameter changes to be more significantly varied in the virtualized system Figure 2 5 shows the result of my test. The virtual environment throughput was smaller than that in the native environment in most of the tests. However, the shown test results had some deviations from what I had expected. The uniquebytes diagram in Figure 2 has a relatively flat (horizontal) graph, which differs from my expectation. This can be resulted from the fact that the size of the uniquebytes is not big enough to force the file system to continuously swapping data in and out. Figure 3, 4, and 5 generally match our expectations. The diagram in Figure 3 supports my predication that as the sizemean increases, the throughput decreases as well as the fact that the virtualized system throughput is more sensitive to size- Mean that the native system. Figure 4 matches with my expectation that as the processnum increases, the throughout decreases. The last graph tells us that as the readfrac increases, the throughput increases in both native and virtualized systems. Also there is a huge jump on the graph of the virtualized system. Since the virtualized system is running on top of the native system, the memory can still exist in the host cache even though the virtualized system clears its own memory. This explains the dramatic increase on graph. Overall, by observing the graphs, I reach the conclusion that this tool is a good benchmark that reflects the performance of HDFS, and the data in this test is not optimal enough to yield more significant results. In order to improve this, I can design a new test with a bigger uniquebytes to reach the decreasing threshold of the throughput as well as increase the default size of the sizemean. 1KB is so small that it could possibly become a factor that dominates the performance (E.g. A default size- Mean of 2KB can take up as long as two hours to run).

4 Figure 2. Results of testing uniquebytes from 64 MB to 2GB in the first experiment. Figure 3. Results of testing sizemean from 1KB to 1MB in the first experiment.

5 Figure 4. Results of testing processnum from 1 to 5 in the first experiment. Figure 5. Results of testing readfrac from 0 to 1 in the first experiment.

6 6. Second Experiment Based on the lessons learned from the first experiment, I conducted a second experiment for an optimal design. In this section, I will discuss the setup and result for the second experiment. 6.1 Setup This time I used two machines running the native and the virtualized systems respectively to speed up the testing process. The hardware configuration for both machines is Inter i5 (2.4 GHz) and the rest of the setting is the same as the first experiment. The design of this experiment is in general the same with the previous one except the default parameter values. The new default parameter values are: uniquebytes 4.5GB, sizemean 1M, processnum 2, readfrac 0.5. I selected this specific set of parameters in order to guarantee a reliable result by imitating a real-life setting. The new test range for the parameters are : uniquebytes (4GB 12GB), sizemean (1KB 2MB), processnum (1-6), and readfrac (0 1). 6.2 Result Surprisingly, in this experiment, the virtualized system outperformed the native system, which deviated from what I expected. This could be possibly attributed to a read fraction of 0.5, a value that is too large for an optimal result. Figure 6 shows another flat graph that is similar to Figure 2. It is possible that the data size is still not big enough to reach the bottleneck. The same reason can be used to explain the flat graph in Figure 7. Figure 8 matches the result from the first experiment. Figure 6. Results of testing uniquebytes from 4 GB to 12 GB in the second experiment.

7 Figure 7. Results of testing sizemean from 32KB to 2MB in the second experiment. Figure 8. Results of test processnum from 1 to 6 in the second experiment.

8 Figure 9. Results of testing readfrac from 0 to 1 in the second experiment. 7. Conclusions and Future Work In this project, I have implemented Chen and Patterson s workload model of I/O system as a better test tool than TestDFSIO to benchmark HDFS. The implementation was tested in both native and virtualized systems. Even though the result I obtained shows that this tool successfully captures the load and concurrency characteristics of HDFS, it also has potentials that are not revealed in this project and yet to be discovered by continuing to explore the parameter space in order to provide synthetic benchmarking. This project inspires me to learn from testing and possibly focus on selfscaling benchmark to find the parameters for optimal performance. Also the current tests were conducted on the pseudo distributive mode of Hadoop and it would be essential to see its performance in full-distributed mode. Moreover, Hadoop s performance in other VM platforms would also be a potential field to study for better performance. 8. References [1] TechNavio. (2013). Global Hadoop-as-a-service Market [Online]. Available: [2] P. M. Chen, D. A. Patterson, "A New Approach to I/O Performance Evaluation--Self-Scaling I/O Benchmarks, Predicted I/O Performance", ACM Transactions on Computer Systems, November [3] J. Dean and S. Ghemawat. "MapReduce: simplified data processing on large clusters." In OSDI '04: Sixth Symposium on Operating System Design and Implementation, pages , [4] The Apache Software Foundation. (2008). Welcome to Apache Hadoop!. [online]. Available:

Hadoop Scheduler w i t h Deadline Constraint

Hadoop Scheduler w i t h Deadline Constraint Hadoop Scheduler w i t h Deadline Constraint Geetha J 1, N UdayBhaskar 2, P ChennaReddy 3,Neha Sniha 4 1,4 Department of Computer Science and Engineering, M S Ramaiah Institute of Technology, Bangalore,

More information

Benchmarking Hadoop & HBase on Violin

Benchmarking Hadoop & HBase on Violin Technical White Paper Report Technical Report Benchmarking Hadoop & HBase on Violin Harnessing Big Data Analytics at the Speed of Memory Version 1.0 Abstract The purpose of benchmarking is to show advantages

More information

MapReduce and Hadoop Distributed File System

MapReduce and Hadoop Distributed File System MapReduce and Hadoop Distributed File System 1 B. RAMAMURTHY Contact: Dr. Bina Ramamurthy CSE Department University at Buffalo (SUNY) bina@buffalo.edu http://www.cse.buffalo.edu/faculty/bina Partially

More information

Energy-Saving Cloud Computing Platform Based On Micro-Embedded System

Energy-Saving Cloud Computing Platform Based On Micro-Embedded System Energy-Saving Cloud Computing Platform Based On Micro-Embedded System Wen-Hsu HSIEH *, San-Peng KAO **, Kuang-Hung TAN **, Jiann-Liang CHEN ** * Department of Computer and Communication, De Lin Institute

More information

Enhancing Dataset Processing in Hadoop YARN Performance for Big Data Applications

Enhancing Dataset Processing in Hadoop YARN Performance for Big Data Applications Enhancing Dataset Processing in Hadoop YARN Performance for Big Data Applications Ahmed Abdulhakim Al-Absi, Dae-Ki Kang and Myong-Jong Kim Abstract In Hadoop MapReduce distributed file system, as the input

More information

Analysis and Modeling of MapReduce s Performance on Hadoop YARN

Analysis and Modeling of MapReduce s Performance on Hadoop YARN Analysis and Modeling of MapReduce s Performance on Hadoop YARN Qiuyi Tang Dept. of Mathematics and Computer Science Denison University tang_j3@denison.edu Dr. Thomas C. Bressoud Dept. of Mathematics and

More information

Hadoop Architecture. Part 1

Hadoop Architecture. Part 1 Hadoop Architecture Part 1 Node, Rack and Cluster: A node is simply a computer, typically non-enterprise, commodity hardware for nodes that contain data. Consider we have Node 1.Then we can add more nodes,

More information

Delivering Quality in Software Performance and Scalability Testing

Delivering Quality in Software Performance and Scalability Testing Delivering Quality in Software Performance and Scalability Testing Abstract Khun Ban, Robert Scott, Kingsum Chow, and Huijun Yan Software and Services Group, Intel Corporation {khun.ban, robert.l.scott,

More information

Introduction to Hadoop

Introduction to Hadoop Introduction to Hadoop 1 What is Hadoop? the big data revolution extracting value from data cloud computing 2 Understanding MapReduce the word count problem more examples MCS 572 Lecture 24 Introduction

More information

GraySort on Apache Spark by Databricks

GraySort on Apache Spark by Databricks GraySort on Apache Spark by Databricks Reynold Xin, Parviz Deyhim, Ali Ghodsi, Xiangrui Meng, Matei Zaharia Databricks Inc. Apache Spark Sorting in Spark Overview Sorting Within a Partition Range Partitioner

More information

Distributed File System. MCSN N. Tonellotto Complements of Distributed Enabling Platforms

Distributed File System. MCSN N. Tonellotto Complements of Distributed Enabling Platforms Distributed File System 1 How do we get data to the workers? NAS Compute Nodes SAN 2 Distributed File System Don t move data to workers move workers to the data! Store data on the local disks of nodes

More information

Benchmark Hadoop and Mars: MapReduce on cluster versus on GPU

Benchmark Hadoop and Mars: MapReduce on cluster versus on GPU Benchmark Hadoop and Mars: MapReduce on cluster versus on GPU Heshan Li, Shaopeng Wang The Johns Hopkins University 3400 N. Charles Street Baltimore, Maryland 21218 {heshanli, shaopeng}@cs.jhu.edu 1 Overview

More information

Cloud Storage. Parallels. Performance Benchmark Results. White Paper. www.parallels.com

Cloud Storage. Parallels. Performance Benchmark Results. White Paper. www.parallels.com Parallels Cloud Storage White Paper Performance Benchmark Results www.parallels.com Table of Contents Executive Summary... 3 Architecture Overview... 3 Key Features... 4 No Special Hardware Requirements...

More information

Introduction to Hadoop

Introduction to Hadoop 1 What is Hadoop? Introduction to Hadoop We are living in an era where large volumes of data are available and the problem is to extract meaning from the data avalanche. The goal of the software tools

More information

Fault Tolerance in Hadoop for Work Migration

Fault Tolerance in Hadoop for Work Migration 1 Fault Tolerance in Hadoop for Work Migration Shivaraman Janakiraman Indiana University Bloomington ABSTRACT Hadoop is a framework that runs applications on large clusters which are built on numerous

More information

A Hybrid Scheduling Approach for Scalable Heterogeneous Hadoop Systems

A Hybrid Scheduling Approach for Scalable Heterogeneous Hadoop Systems A Hybrid Scheduling Approach for Scalable Heterogeneous Hadoop Systems Aysan Rasooli Department of Computing and Software McMaster University Hamilton, Canada Email: rasooa@mcmaster.ca Douglas G. Down

More information

Big Fast Data Hadoop acceleration with Flash. June 2013

Big Fast Data Hadoop acceleration with Flash. June 2013 Big Fast Data Hadoop acceleration with Flash June 2013 Agenda The Big Data Problem What is Hadoop Hadoop and Flash The Nytro Solution Test Results The Big Data Problem Big Data Output Facebook Traditional

More information

Mobile Cloud Computing for Data-Intensive Applications

Mobile Cloud Computing for Data-Intensive Applications Mobile Cloud Computing for Data-Intensive Applications Senior Thesis Final Report Vincent Teo, vct@andrew.cmu.edu Advisor: Professor Priya Narasimhan, priya@cs.cmu.edu Abstract The computational and storage

More information

Performance Characteristics of VMFS and RDM VMware ESX Server 3.0.1

Performance Characteristics of VMFS and RDM VMware ESX Server 3.0.1 Performance Study Performance Characteristics of and RDM VMware ESX Server 3.0.1 VMware ESX Server offers three choices for managing disk access in a virtual machine VMware Virtual Machine File System

More information

A REVIEW PAPER ON THE HADOOP DISTRIBUTED FILE SYSTEM

A REVIEW PAPER ON THE HADOOP DISTRIBUTED FILE SYSTEM A REVIEW PAPER ON THE HADOOP DISTRIBUTED FILE SYSTEM Sneha D.Borkar 1, Prof.Chaitali S.Surtakar 2 Student of B.E., Information Technology, J.D.I.E.T, sborkar95@gmail.com Assistant Professor, Information

More information

The Comprehensive Performance Rating for Hadoop Clusters on Cloud Computing Platform

The Comprehensive Performance Rating for Hadoop Clusters on Cloud Computing Platform The Comprehensive Performance Rating for Hadoop Clusters on Cloud Computing Platform Fong-Hao Liu, Ya-Ruei Liou, Hsiang-Fu Lo, Ko-Chin Chang, and Wei-Tsong Lee Abstract Virtualization platform solutions

More information

Survey of the Benchmark Systems and Testing Frameworks For Tachyon-Perf

Survey of the Benchmark Systems and Testing Frameworks For Tachyon-Perf Survey of the Benchmark Systems and Testing Frameworks For Tachyon-Perf Rong Gu,Qianhao Dong 2014/09/05 0. Introduction As we want to have a performance framework for Tachyon, we need to consider two aspects

More information

Marvell DragonFly Virtual Storage Accelerator Performance Benchmarks

Marvell DragonFly Virtual Storage Accelerator Performance Benchmarks PERFORMANCE BENCHMARKS PAPER Marvell DragonFly Virtual Storage Accelerator Performance Benchmarks Arvind Pruthi Senior Staff Manager Marvell April 2011 www.marvell.com Overview In today s virtualized data

More information

A Framework for Performance Analysis and Tuning in Hadoop Based Clusters

A Framework for Performance Analysis and Tuning in Hadoop Based Clusters A Framework for Performance Analysis and Tuning in Hadoop Based Clusters Garvit Bansal Anshul Gupta Utkarsh Pyne LNMIIT, Jaipur, India Email: [garvit.bansal anshul.gupta utkarsh.pyne] @lnmiit.ac.in Manish

More information

Maximizing Hadoop Performance and Storage Capacity with AltraHD TM

Maximizing Hadoop Performance and Storage Capacity with AltraHD TM Maximizing Hadoop Performance and Storage Capacity with AltraHD TM Executive Summary The explosion of internet data, driven in large part by the growth of more and more powerful mobile devices, has created

More information

GraySort and MinuteSort at Yahoo on Hadoop 0.23

GraySort and MinuteSort at Yahoo on Hadoop 0.23 GraySort and at Yahoo on Hadoop.23 Thomas Graves Yahoo! May, 213 The Apache Hadoop[1] software library is an open source framework that allows for the distributed processing of large data sets across clusters

More information

BENCHMARKING CLOUD DATABASES CASE STUDY on HBASE, HADOOP and CASSANDRA USING YCSB

BENCHMARKING CLOUD DATABASES CASE STUDY on HBASE, HADOOP and CASSANDRA USING YCSB BENCHMARKING CLOUD DATABASES CASE STUDY on HBASE, HADOOP and CASSANDRA USING YCSB Planet Size Data!? Gartner s 10 key IT trends for 2012 unstructured data will grow some 80% over the course of the next

More information

MapReduce and Hadoop Distributed File System V I J A Y R A O

MapReduce and Hadoop Distributed File System V I J A Y R A O MapReduce and Hadoop Distributed File System 1 V I J A Y R A O The Context: Big-data Man on the moon with 32KB (1969); my laptop had 2GB RAM (2009) Google collects 270PB data in a month (2007), 20000PB

More information

GeoGrid Project and Experiences with Hadoop

GeoGrid Project and Experiences with Hadoop GeoGrid Project and Experiences with Hadoop Gong Zhang and Ling Liu Distributed Data Intensive Systems Lab (DiSL) Center for Experimental Computer Systems Research (CERCS) Georgia Institute of Technology

More information

BIG DATA USING HADOOP

BIG DATA USING HADOOP + Breakaway Session By Johnson Iyilade, Ph.D. University of Saskatchewan, Canada 23-July, 2015 BIG DATA USING HADOOP + Outline n Framing the Problem Hadoop Solves n Meet Hadoop n Storage with HDFS n Data

More information

Big Data and Apache Hadoop s MapReduce

Big Data and Apache Hadoop s MapReduce Big Data and Apache Hadoop s MapReduce Michael Hahsler Computer Science and Engineering Southern Methodist University January 23, 2012 Michael Hahsler (SMU/CSE) Hadoop/MapReduce January 23, 2012 1 / 23

More information

Chapter 7. Using Hadoop Cluster and MapReduce

Chapter 7. Using Hadoop Cluster and MapReduce Chapter 7 Using Hadoop Cluster and MapReduce Modeling and Prototyping of RMS for QoS Oriented Grid Page 152 7. Using Hadoop Cluster and MapReduce for Big Data Problems The size of the databases used in

More information

Performance Analysis: Benchmarking Public Clouds

Performance Analysis: Benchmarking Public Clouds Performance Analysis: Benchmarking Public Clouds Performance comparison of web server and database VMs on Internap AgileCLOUD and Amazon Web Services By Cloud Spectator March 215 PERFORMANCE REPORT WEB

More information

Energy Efficient MapReduce

Energy Efficient MapReduce Energy Efficient MapReduce Motivation: Energy consumption is an important aspect of datacenters efficiency, the total power consumption in the united states has doubled from 2000 to 2005, representing

More information

A Case for Flash Memory SSD in Hadoop Applications

A Case for Flash Memory SSD in Hadoop Applications A Case for Flash Memory SSD in Hadoop Applications Seok-Hoon Kang, Dong-Hyun Koo, Woon-Hak Kang and Sang-Won Lee Dept of Computer Engineering, Sungkyunkwan University, Korea x860221@gmail.com, smwindy@naver.com,

More information

Use of Hadoop File System for Nuclear Physics Analyses in STAR

Use of Hadoop File System for Nuclear Physics Analyses in STAR 1 Use of Hadoop File System for Nuclear Physics Analyses in STAR EVAN SANGALINE UC DAVIS Motivations 2 Data storage a key component of analysis requirements Transmission and storage across diverse resources

More information

WHITE PAPER Optimizing Virtual Platform Disk Performance

WHITE PAPER Optimizing Virtual Platform Disk Performance WHITE PAPER Optimizing Virtual Platform Disk Performance Think Faster. Visit us at Condusiv.com Optimizing Virtual Platform Disk Performance 1 The intensified demand for IT network efficiency and lower

More information

BIG DATA TRENDS AND TECHNOLOGIES

BIG DATA TRENDS AND TECHNOLOGIES BIG DATA TRENDS AND TECHNOLOGIES THE WORLD OF DATA IS CHANGING Cloud WHAT IS BIG DATA? Big data are datasets that grow so large that they become awkward to work with using onhand database management tools.

More information

Application Development. A Paradigm Shift

Application Development. A Paradigm Shift Application Development for the Cloud: A Paradigm Shift Ramesh Rangachar Intelsat t 2012 by Intelsat. t Published by The Aerospace Corporation with permission. New 2007 Template - 1 Motivation for the

More information

DIABLO TECHNOLOGIES MEMORY CHANNEL STORAGE AND VMWARE VIRTUAL SAN : VDI ACCELERATION

DIABLO TECHNOLOGIES MEMORY CHANNEL STORAGE AND VMWARE VIRTUAL SAN : VDI ACCELERATION DIABLO TECHNOLOGIES MEMORY CHANNEL STORAGE AND VMWARE VIRTUAL SAN : VDI ACCELERATION A DIABLO WHITE PAPER AUGUST 2014 Ricky Trigalo Director of Business Development Virtualization, Diablo Technologies

More information

HP Z Turbo Drive PCIe SSD

HP Z Turbo Drive PCIe SSD Performance Evaluation of HP Z Turbo Drive PCIe SSD Powered by Samsung XP941 technology Evaluation Conducted Independently by: Hamid Taghavi Senior Technical Consultant June 2014 Sponsored by: P a g e

More information

Analysis and Optimization of Massive Data Processing on High Performance Computing Architecture

Analysis and Optimization of Massive Data Processing on High Performance Computing Architecture Analysis and Optimization of Massive Data Processing on High Performance Computing Architecture He Huang, Shanshan Li, Xiaodong Yi, Feng Zhang, Xiangke Liao and Pan Dong School of Computer Science National

More information

A Study on Workload Imbalance Issues in Data Intensive Distributed Computing

A Study on Workload Imbalance Issues in Data Intensive Distributed Computing A Study on Workload Imbalance Issues in Data Intensive Distributed Computing Sven Groot 1, Kazuo Goda 1, and Masaru Kitsuregawa 1 University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Japan Abstract.

More information

Distributed File Systems

Distributed File Systems Distributed File Systems Paul Krzyzanowski Rutgers University October 28, 2012 1 Introduction The classic network file systems we examined, NFS, CIFS, AFS, Coda, were designed as client-server applications.

More information

Virtuoso and Database Scalability

Virtuoso and Database Scalability Virtuoso and Database Scalability By Orri Erling Table of Contents Abstract Metrics Results Transaction Throughput Initializing 40 warehouses Serial Read Test Conditions Analysis Working Set Effect of

More information

Storage I/O Performance on VMware vsphere

Storage I/O Performance on VMware vsphere Storage I/O Performance on VMware vsphere 5.1 over 16 Gigabit Fibre Channel Performance Study TECHNICAL WHITE PAPER Table of Contents Introduction... 3 Executive Summary... 3 Setup... 3 Workload... 4 Results...

More information

Talend Real-Time Big Data Sandbox. Big Data Insights Cookbook

Talend Real-Time Big Data Sandbox. Big Data Insights Cookbook Talend Real-Time Big Data Talend Real-Time Big Data Overview of Real-time Big Data Pre-requisites to run Setup & Talend License Talend Real-Time Big Data Big Data Setup & About this cookbook What is the

More information

HP ProLiant Gen8 vs Gen9 Server Blades on Data Warehouse Workloads

HP ProLiant Gen8 vs Gen9 Server Blades on Data Warehouse Workloads HP ProLiant Gen8 vs Gen9 Server Blades on Data Warehouse Workloads Gen9 Servers give more performance per dollar for your investment. Executive Summary Information Technology (IT) organizations face increasing

More information

Experiences with Lustre* and Hadoop*

Experiences with Lustre* and Hadoop* Experiences with Lustre* and Hadoop* Gabriele Paciucci (Intel) June, 2014 Intel * Some Con fidential name Do Not Forward and brands may be claimed as the property of others. Agenda Overview Intel Enterprise

More information

The Performance Characteristics of MapReduce Applications on Scalable Clusters

The Performance Characteristics of MapReduce Applications on Scalable Clusters The Performance Characteristics of MapReduce Applications on Scalable Clusters Kenneth Wottrich Denison University Granville, OH 43023 wottri_k1@denison.edu ABSTRACT Many cluster owners and operators have

More information

Reducer Load Balancing and Lazy Initialization in Map Reduce Environment S.Mohanapriya, P.Natesan

Reducer Load Balancing and Lazy Initialization in Map Reduce Environment S.Mohanapriya, P.Natesan Reducer Load Balancing and Lazy Initialization in Map Reduce Environment S.Mohanapriya, P.Natesan Abstract Big Data is revolutionizing 21st-century with increasingly huge amounts of data to store and be

More information

CSE-E5430 Scalable Cloud Computing Lecture 2

CSE-E5430 Scalable Cloud Computing Lecture 2 CSE-E5430 Scalable Cloud Computing Lecture 2 Keijo Heljanko Department of Computer Science School of Science Aalto University keijo.heljanko@aalto.fi 14.9-2015 1/36 Google MapReduce A scalable batch processing

More information

Journal of science STUDY ON REPLICA MANAGEMENT AND HIGH AVAILABILITY IN HADOOP DISTRIBUTED FILE SYSTEM (HDFS)

Journal of science STUDY ON REPLICA MANAGEMENT AND HIGH AVAILABILITY IN HADOOP DISTRIBUTED FILE SYSTEM (HDFS) Journal of science e ISSN 2277-3290 Print ISSN 2277-3282 Information Technology www.journalofscience.net STUDY ON REPLICA MANAGEMENT AND HIGH AVAILABILITY IN HADOOP DISTRIBUTED FILE SYSTEM (HDFS) S. Chandra

More information

Performance Analysis of Mixed Distributed Filesystem Workloads

Performance Analysis of Mixed Distributed Filesystem Workloads Performance Analysis of Mixed Distributed Filesystem Workloads Esteban Molina-Estolano, Maya Gokhale, Carlos Maltzahn, John May, John Bent, Scott Brandt Motivation Hadoop-tailored filesystems (e.g. CloudStore)

More information

RAID 5 rebuild performance in ProLiant

RAID 5 rebuild performance in ProLiant RAID 5 rebuild performance in ProLiant technology brief Abstract... 2 Overview of the RAID 5 rebuild process... 2 Estimating the mean-time-to-failure (MTTF)... 3 Factors affecting RAID 5 array rebuild

More information

III Big Data Technologies

III Big Data Technologies III Big Data Technologies Today, new technologies make it possible to realize value from Big Data. Big data technologies can replace highly customized, expensive legacy systems with a standard solution

More information

Computing in clouds: Where we come from, Where we are, What we can, Where we go

Computing in clouds: Where we come from, Where we are, What we can, Where we go Computing in clouds: Where we come from, Where we are, What we can, Where we go Luc Bougé ENS Cachan/Rennes, IRISA, INRIA Biogenouest With help from many colleagues: Gabriel Antoniu, Guillaume Pierre,

More information

Department of Computer Science University of Cyprus EPL646 Advanced Topics in Databases. Lecture 14

Department of Computer Science University of Cyprus EPL646 Advanced Topics in Databases. Lecture 14 Department of Computer Science University of Cyprus EPL646 Advanced Topics in Databases Lecture 14 Big Data Management IV: Big-data Infrastructures (Background, IO, From NFS to HFDS) Chapter 14-15: Abideboul

More information

Exploring RAID Configurations

Exploring RAID Configurations Exploring RAID Configurations J. Ryan Fishel Florida State University August 6, 2008 Abstract To address the limits of today s slow mechanical disks, we explored a number of data layouts to improve RAID

More information

Open Cloud System. (Integration of Eucalyptus, Hadoop and AppScale into deployment of University Private Cloud)

Open Cloud System. (Integration of Eucalyptus, Hadoop and AppScale into deployment of University Private Cloud) Open Cloud System (Integration of Eucalyptus, Hadoop and into deployment of University Private Cloud) Thinn Thu Naing University of Computer Studies, Yangon 25 th October 2011 Open Cloud System University

More information

Improving MapReduce Performance in Heterogeneous Environments

Improving MapReduce Performance in Heterogeneous Environments UC Berkeley Improving MapReduce Performance in Heterogeneous Environments Matei Zaharia, Andy Konwinski, Anthony Joseph, Randy Katz, Ion Stoica University of California at Berkeley Motivation 1. MapReduce

More information

Apache Hadoop. Alexandru Costan

Apache Hadoop. Alexandru Costan 1 Apache Hadoop Alexandru Costan Big Data Landscape No one-size-fits-all solution: SQL, NoSQL, MapReduce, No standard, except Hadoop 2 Outline What is Hadoop? Who uses it? Architecture HDFS MapReduce Open

More information

Accelerating and Simplifying Apache

Accelerating and Simplifying Apache Accelerating and Simplifying Apache Hadoop with Panasas ActiveStor White paper NOvember 2012 1.888.PANASAS www.panasas.com Executive Overview The technology requirements for big data vary significantly

More information

Benchmarking Cassandra on Violin

Benchmarking Cassandra on Violin Technical White Paper Report Technical Report Benchmarking Cassandra on Violin Accelerating Cassandra Performance and Reducing Read Latency With Violin Memory Flash-based Storage Arrays Version 1.0 Abstract

More information

Scalable Multiple NameNodes Hadoop Cloud Storage System

Scalable Multiple NameNodes Hadoop Cloud Storage System Vol.8, No.1 (2015), pp.105-110 http://dx.doi.org/10.14257/ijdta.2015.8.1.12 Scalable Multiple NameNodes Hadoop Cloud Storage System Kun Bi 1 and Dezhi Han 1,2 1 College of Information Engineering, Shanghai

More information

marlabs driving digital agility WHITEPAPER Big Data and Hadoop

marlabs driving digital agility WHITEPAPER Big Data and Hadoop marlabs driving digital agility WHITEPAPER Big Data and Hadoop Abstract This paper explains the significance of Hadoop, an emerging yet rapidly growing technology. The prime goal of this paper is to unveil

More information

Welcome to the unit of Hadoop Fundamentals on Hadoop architecture. I will begin with a terminology review and then cover the major components

Welcome to the unit of Hadoop Fundamentals on Hadoop architecture. I will begin with a terminology review and then cover the major components Welcome to the unit of Hadoop Fundamentals on Hadoop architecture. I will begin with a terminology review and then cover the major components of Hadoop. We will see what types of nodes can exist in a Hadoop

More information

BlobSeer: Towards efficient data storage management on large-scale, distributed systems

BlobSeer: Towards efficient data storage management on large-scale, distributed systems : Towards efficient data storage management on large-scale, distributed systems Bogdan Nicolae University of Rennes 1, France KerData Team, INRIA Rennes Bretagne-Atlantique PhD Advisors: Gabriel Antoniu

More information

A METHODOLOGY FOR IDENTIFYING THE RELATIONSHIP BETWEEN PERFORMANCE FACTORS FOR CLOUD COMPUTING APPLICATIONS

A METHODOLOGY FOR IDENTIFYING THE RELATIONSHIP BETWEEN PERFORMANCE FACTORS FOR CLOUD COMPUTING APPLICATIONS A METHODOLOGY FOR IDENTIFYING THE RELATIONSHIP BETWEEN PERFORMANCE FACTORS FOR CLOUD COMPUTING APPLICATIONS Luis Bautista 1, 2, Alain April 2, Alain Abran 2 1 Department of Electronic Systems, Autonomous

More information

Linux Filesystem Performance Comparison for OLTP with Ext2, Ext3, Raw, and OCFS on Direct-Attached Disks using Oracle 9i Release 2

Linux Filesystem Performance Comparison for OLTP with Ext2, Ext3, Raw, and OCFS on Direct-Attached Disks using Oracle 9i Release 2 Linux Filesystem Performance Comparison for OLTP with Ext2, Ext3, Raw, and OCFS on Direct-Attached Disks using Oracle 9i Release 2 An Oracle White Paper January 2004 Linux Filesystem Performance Comparison

More information

How In-Memory Data Grids Can Analyze Fast-Changing Data in Real Time

How In-Memory Data Grids Can Analyze Fast-Changing Data in Real Time SCALEOUT SOFTWARE How In-Memory Data Grids Can Analyze Fast-Changing Data in Real Time by Dr. William Bain and Dr. Mikhail Sobolev, ScaleOut Software, Inc. 2012 ScaleOut Software, Inc. 12/27/2012 T wenty-first

More information

Scalability Factors of JMeter In Performance Testing Projects

Scalability Factors of JMeter In Performance Testing Projects Scalability Factors of JMeter In Performance Testing Projects Title Scalability Factors for JMeter In Performance Testing Projects Conference STEP-IN Conference Performance Testing 2008, PUNE Author(s)

More information

Hadoop on OpenStack Cloud. Dmitry Mescheryakov Software Engineer, @MirantisIT

Hadoop on OpenStack Cloud. Dmitry Mescheryakov Software Engineer, @MirantisIT Hadoop on OpenStack Cloud Dmitry Mescheryakov Software Engineer, @MirantisIT Agenda OpenStack Sahara Demo Hadoop Performance on Cloud Conclusion OpenStack Open source cloud computing platform 17,209 commits

More information

Performance And Scalability In Oracle9i And SQL Server 2000

Performance And Scalability In Oracle9i And SQL Server 2000 Performance And Scalability In Oracle9i And SQL Server 2000 Presented By : Phathisile Sibanda Supervisor : John Ebden 1 Presentation Overview Project Objectives Motivation -Why performance & Scalability

More information

Performance Comparison of Intel Enterprise Edition for Lustre* software and HDFS for MapReduce Applications

Performance Comparison of Intel Enterprise Edition for Lustre* software and HDFS for MapReduce Applications Performance Comparison of Intel Enterprise Edition for Lustre software and HDFS for MapReduce Applications Rekha Singhal, Gabriele Pacciucci and Mukesh Gangadhar 2 Hadoop Introduc-on Open source MapReduce

More information

Performance Comparison of SQL based Big Data Analytics with Lustre and HDFS file systems

Performance Comparison of SQL based Big Data Analytics with Lustre and HDFS file systems Performance Comparison of SQL based Big Data Analytics with Lustre and HDFS file systems Rekha Singhal and Gabriele Pacciucci * Other names and brands may be claimed as the property of others. Lustre File

More information

Introducing EEMBC Cloud and Big Data Server Benchmarks

Introducing EEMBC Cloud and Big Data Server Benchmarks Introducing EEMBC Cloud and Big Data Server Benchmarks Quick Background: Industry-Standard Benchmarks for the Embedded Industry EEMBC formed in 1997 as non-profit consortium Defining and developing application-specific

More information

IOmark- VDI. Nimbus Data Gemini Test Report: VDI- 130906- a Test Report Date: 6, September 2013. www.iomark.org

IOmark- VDI. Nimbus Data Gemini Test Report: VDI- 130906- a Test Report Date: 6, September 2013. www.iomark.org IOmark- VDI Nimbus Data Gemini Test Report: VDI- 130906- a Test Copyright 2010-2013 Evaluator Group, Inc. All rights reserved. IOmark- VDI, IOmark- VDI, VDI- IOmark, and IOmark are trademarks of Evaluator

More information

Index Terms : Load rebalance, distributed file systems, clouds, movement cost, load imbalance, chunk.

Index Terms : Load rebalance, distributed file systems, clouds, movement cost, load imbalance, chunk. Load Rebalancing for Distributed File Systems in Clouds. Smita Salunkhe, S. S. Sannakki Department of Computer Science and Engineering KLS Gogte Institute of Technology, Belgaum, Karnataka, India Affiliated

More information

12th WSEAS International Conference on COMPUTERS, Heraklion, Greece, July 23-25, 2008

12th WSEAS International Conference on COMPUTERS, Heraklion, Greece, July 23-25, 2008 Specification and Implementation of Dynamic Web Site Benchmark In Telecommunication Area Prof. Dr. EBADA SARHAN* Prof. Dr. ATIF GHALWASH* MOHAMED KHAFAGY** * Computer Science Department, Faculty of Computers

More information

SOLUTION BRIEF: SLCM R12.8 PERFORMANCE TEST RESULTS JANUARY, 2013. Submit and Approval Phase Results

SOLUTION BRIEF: SLCM R12.8 PERFORMANCE TEST RESULTS JANUARY, 2013. Submit and Approval Phase Results SOLUTION BRIEF: SLCM R12.8 PERFORMANCE TEST RESULTS JANUARY, 2013 Submit and Approval Phase Results Table of Contents Executive Summary 3 Test Environment 4 Server Topology 4 CA Service Catalog Settings

More information

Comparison of Windows IaaS Environments

Comparison of Windows IaaS Environments Comparison of Windows IaaS Environments Comparison of Amazon Web Services, Expedient, Microsoft, and Rackspace Public Clouds January 5, 215 TABLE OF CONTENTS Executive Summary 2 vcpu Performance Summary

More information

Resource Scalability for Efficient Parallel Processing in Cloud

Resource Scalability for Efficient Parallel Processing in Cloud Resource Scalability for Efficient Parallel Processing in Cloud ABSTRACT Govinda.K #1, Abirami.M #2, Divya Mercy Silva.J #3 #1 SCSE, VIT University #2 SITE, VIT University #3 SITE, VIT University In the

More information

Data Center Specific Thermal and Energy Saving Techniques

Data Center Specific Thermal and Energy Saving Techniques Data Center Specific Thermal and Energy Saving Techniques Tausif Muzaffar and Xiao Qin Department of Computer Science and Software Engineering Auburn University 1 Big Data 2 Data Centers In 2013, there

More information

Overhead and Performance Impact when Using Full Drive Encryption with HP ProtectTools and SSD

Overhead and Performance Impact when Using Full Drive Encryption with HP ProtectTools and SSD After some weeks of waiting, I received my new HP EliteBook 8440p with the following parameters: Processor type Intel Core i7-620m Processor (2.66 GHz, 4 MB L3 cache) Operating system installed Windows

More information

Comparative analysis of mapreduce job by keeping data constant and varying cluster size technique

Comparative analysis of mapreduce job by keeping data constant and varying cluster size technique Comparative analysis of mapreduce job by keeping data constant and varying cluster size technique Mahesh Maurya a, Sunita Mahajan b * a Research Scholar, JJT University, MPSTME, Mumbai, India,maheshkmaurya@yahoo.co.in

More information

MapReduce on GPUs. Amit Sabne, Ahmad Mujahid Mohammed Razip, Kun Xu

MapReduce on GPUs. Amit Sabne, Ahmad Mujahid Mohammed Razip, Kun Xu 1 MapReduce on GPUs Amit Sabne, Ahmad Mujahid Mohammed Razip, Kun Xu 2 MapReduce MAP Shuffle Reduce 3 Hadoop Open-source MapReduce framework from Apache, written in Java Used by Yahoo!, Facebook, Ebay,

More information

NoSQL Data Base Basics

NoSQL Data Base Basics NoSQL Data Base Basics Course Notes in Transparency Format Cloud Computing MIRI (CLC-MIRI) UPC Master in Innovation & Research in Informatics Spring- 2013 Jordi Torres, UPC - BSC www.jorditorres.eu HDFS

More information

The Hadoop Framework

The Hadoop Framework The Hadoop Framework Nils Braden University of Applied Sciences Gießen-Friedberg Wiesenstraße 14 35390 Gießen nils.braden@mni.fh-giessen.de Abstract. The Hadoop Framework offers an approach to large-scale

More information

Hadoop IST 734 SS CHUNG

Hadoop IST 734 SS CHUNG Hadoop IST 734 SS CHUNG Introduction What is Big Data?? Bulk Amount Unstructured Lots of Applications which need to handle huge amount of data (in terms of 500+ TB per day) If a regular machine need to

More information

Industry 8Gb / 16Gb Fibre Channel HBA Evaluation

Industry 8Gb / 16Gb Fibre Channel HBA Evaluation Industry 8Gb / 16Gb Fibre Channel HBA Evaluation Evaluation report prepared under contract with QLogic Executive Summary Explosive growth in the complexity and amount of data of today s datacenter environments

More information

Big Data and Natural Language: Extracting Insight From Text

Big Data and Natural Language: Extracting Insight From Text An Oracle White Paper October 2012 Big Data and Natural Language: Extracting Insight From Text Table of Contents Executive Overview... 3 Introduction... 3 Oracle Big Data Appliance... 4 Synthesys... 5

More information

Snapshots in Hadoop Distributed File System

Snapshots in Hadoop Distributed File System Snapshots in Hadoop Distributed File System Sameer Agarwal UC Berkeley Dhruba Borthakur Facebook Inc. Ion Stoica UC Berkeley Abstract The ability to take snapshots is an essential functionality of any

More information

Pepper: An Elastic Web Server Farm for Cloud based on Hadoop. Subramaniam Krishnan, Jean Christophe Counio Yahoo! Inc. MAPRED 1 st December 2010

Pepper: An Elastic Web Server Farm for Cloud based on Hadoop. Subramaniam Krishnan, Jean Christophe Counio Yahoo! Inc. MAPRED 1 st December 2010 Pepper: An Elastic Web Server Farm for Cloud based on Hadoop Subramaniam Krishnan, Jean Christophe Counio. MAPRED 1 st December 2010 Agenda Motivation Design Features Applications Evaluation Conclusion

More information

HiBench Installation. Sunil Raiyani, Jayam Modi

HiBench Installation. Sunil Raiyani, Jayam Modi HiBench Installation Sunil Raiyani, Jayam Modi Last Updated: May 23, 2014 CONTENTS Contents 1 Introduction 1 2 Installation 1 3 HiBench Benchmarks[3] 1 3.1 Micro Benchmarks..............................

More information

MyCloudLab: An Interactive Web-based Management System for Cloud Computing Administration

MyCloudLab: An Interactive Web-based Management System for Cloud Computing Administration MyCloudLab: An Interactive Web-based Management System for Cloud Computing Administration Hoi-Wan Chan 1, Min Xu 2, Chung-Pan Tang 1, Patrick P. C. Lee 1 & Tsz-Yeung Wong 1, 1 Department of Computer Science

More information

Leveraging BlobSeer to boost up the deployment and execution of Hadoop applications in Nimbus cloud environments on Grid 5000

Leveraging BlobSeer to boost up the deployment and execution of Hadoop applications in Nimbus cloud environments on Grid 5000 Leveraging BlobSeer to boost up the deployment and execution of Hadoop applications in Nimbus cloud environments on Grid 5000 Alexandra Carpen-Amarie Diana Moise Bogdan Nicolae KerData Team, INRIA Outline

More information

Virtual Machine Based Resource Allocation For Cloud Computing Environment

Virtual Machine Based Resource Allocation For Cloud Computing Environment Virtual Machine Based Resource Allocation For Cloud Computing Environment D.Udaya Sree M.Tech (CSE) Department Of CSE SVCET,Chittoor. Andra Pradesh, India Dr.J.Janet Head of Department Department of CSE

More information

Scalability Results. Select the right hardware configuration for your organization to optimize performance

Scalability Results. Select the right hardware configuration for your organization to optimize performance Scalability Results Select the right hardware configuration for your organization to optimize performance Table of Contents Introduction... 1 Scalability... 2 Definition... 2 CPU and Memory Usage... 2

More information

White Paper. Recording Server Virtualization

White Paper. Recording Server Virtualization White Paper Recording Server Virtualization Prepared by: Mike Sherwood, Senior Solutions Engineer Milestone Systems 23 March 2011 Table of Contents Introduction... 3 Target audience and white paper purpose...

More information