Cluster Computing by Pritam Bhansali

download Cluster Computing by Pritam Bhansali

of 17

Transcript of Cluster Computing by Pritam Bhansali

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    1/17

    SEMINAR REPORT

    ON

    CLUSTER COMPUTING

    Submitted to

    Rashtrasant Tukadoji Maharaj Nagpur University,

    NAGPUR

    In partial fulfillment of the requirement of

    M. Sc.- I (Computer Science) Semester-II Examination

    Submitted by

    PRITAM S. BHASALI

    Under the Supervision of

    Dr. S.R.Pande

    Associate professor and Head

    (Department of Computer science)

    DEPARTMENT OF COMPUTER SCIENCE

    Shri Shivaji Education Society Amravatis

    SCIENCE COLLEGECongress Nagar, Nagpur-12.

    2012-2013

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    2/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 1

    INDEX

    Serial number Topic Page No

    1 Abstract 2

    2 Introduction 3

    3 Objectives 5

    4 History 6

    5 How cluster computers works 8

    6 Categorization of clusters 10

    7 Features of cluster computer 11

    8 Advantages and Disadvantages 12

    9 Implementation 13

    10 Future scope 15

    11 Conclusion 16

    12 References 17

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    3/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 2

    Abstract

    Cluster is nothing but group of machines combined to perform the large data

    manipulation by treating that group of machine a single machine by using a large software

    system to control the entire hardware of the machine which is commonly called as Cluster

    Management Software. The whole idea is made and still evolving because to make the

    management of larger amount of data which can be in zeta byte or petabytes and so on.

    Hence the cluster computing encapsulates a combination of hardware platforms and

    software platforms in a single system to manage and store these all types of data that

    includes videos, audios, text formats, various types of related graphics etc. and gives the

    faster access to all of these data. We have also mentioned general categorization of clusters

    such as Load balancing clusters,High-availability (HA) ,High- Performance(HP).

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    4/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 3

    Introduction

    Now the world is moving with a very fast speed towards a new eras in technology, and

    other fields .During such effective movements different types of people like employs,

    scientists etc in addition with these different types of research institutes, organizations and

    private companies are working day and night. While they are working a tons of data are

    getting collected which is very important for them. But this data need to be stored and

    managed properly, hence for storing, effective management and fast access the clusters of

    the machines are formed. A cluster is a local computing system comprising a set of

    independent computers and a network interconnecting them. A cluster is local in that all of its

    component subsystems are supervised within a single administrative domain, usually residingin a single room and managed as a single computer system.

    Clusters are usually deployed to improve performance and/or availability over that

    provided by a single computer, while typically being much more cost-effective than single

    computers of comparable speed or availability.

    A cluster can be either homogeneous (only using PCs) or heterogeneous (using PC and

    workstations).S

    A typical clusters are shown in following images

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    5/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 4

    Objectives

    The major objective in the cluster is utilizing a group of processingnodes so as to complete the assigned job in a minimum amount of time

    by working cooperatively.

    Cluster computing can be used for load balancing as well as for highavailability.

    The main and important strategy to achieve such objective is bytransferring the extra loads from busy nodes to idle nodes.

    The main and important strategy to achieve such objective is bytransferring the extra loads from busy nodes to idle nodes.

    Cluster technologies have been developed that allow multiple low costcomputers to work in a coordinated fashion to process applications.

    Cluster operating systems divide the tasks amongst the available servers. Clusters of systems or workstations, on the other hand, connect a group

    of systems together to jointly share a critically demanding computational

    task.

    These all subsystems and machines are connected through the network technology like

    LAN(local area networks).

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    6/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 5

    History

    Greg Pfister is a computer scientist and worked a lot for cluster computing,

    according to him clusters were not invented by any specific vendor but by customers. The

    formal engineering basis of cluster computing as a means of doing parallel work of any sort

    was arguably invented by Gene Amdahl of IBM who published some papers on related

    topics about cluster computing in 1967.

    The first commercial clustering product was ARCnet(Attached Resource ComputerNETwork), developed by Datapoint in 1977. Then Digital Equipment

    Corporation(DEC) released theirVAXcluster product in 1984 for the VAX/VMS operating

    system then after some timeIBM S/390 Parallel Sysplex was unveiled .

    Consequently the history of early computer clusters is more or less directly tied

    into the history of early networks, as one of the primary motivations for the development of a

    network was to link computing resources, creating a de facto (i.e. in practice or actuality, but

    not officially established. It is commonly used in contrast to dejure (which means concerning

    the law) computer cluster.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    7/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 6

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    8/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 7

    How cluster computers works

    In the above arrangement of clusters the machines are arranged through the common high

    speed bus in a manner that they can communicate with each other and these can attached to a

    common PC or Workstation to manage these all.

    When the clusters of all these machines are formed then it may be treated as a single

    system. This is managed through a commonly known software is Cluster Management

    Software.

    A simple High speed LAN can be used to connect these machines through which these

    machines can be accessed via web or this machine can access the web.

    But in the large organizations the clusters that are formed can be eitherHomogeneous or

    Heterogeneous.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    9/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 8

    Homogeneous clusters are those in which either PCs or Workstations will be used but

    not both are used.

    Heterogeneous clusters are those in which both PCs and Workstations are used.But in the large organizations clusters can be made of more than one storage devices attached

    to a single workstation and number of storage devices can be totally equal to or more then

    petabytes or more.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    10/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 9

    Categorization of clusters

    High-availability (HA)igh availability clusters are generally made in a manner that data loss at the time of

    failure can be avoided. This is done by using an extra backup storages for each of the

    storage machine.

    So that at the time of failure data recovery is easily done from that backup machine.

    Load-balancingLoad balancing is handled by using faster LAN setups and by reducing extra

    hardware.

    High- Performance(HP)High processing clusters generally includes applications like data mining, parallel

    processing, weather modeling, etc.

    The best example of HP clusters is the NASAs Beowulf cluster

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    11/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 10

    Features of cluster computer

    Now clusters became very famous and the are widely needed by various organizations due

    to their effective features which are as follows-

    Easy to deployThe cluster computing system is very easy to deploy. In this system software is installed

    as well as configured automatically. Using web interface, the cluster nodes can be easily

    added and managed and hence, reduces efforts and saves time.

    Easy to manageThe system is very easy to manage as there is no need to edit shell scripts or XML templates.

    It changes node group definitions and maintain several software versions with ease. It takes

    the risk out of software and hardware upgrades as it supports them without upgrading the

    installer node.

    FlexibleAs the cluster computing is an open system, it is very flexible. It supports real-world

    topologies and synchronizes the cluster files without re-installation. The system easily utilizes

    the power of advanced GPUs (Graphic Processing Units) for general HPC calculations. It can

    change software configurations at any time.

    ExpandableIt is very east to add new, future hardware models and cluster node at any time. It is easily

    upgrade to Platform LSF which has proven scalability to 10,000+ CPUs. The commercial

    add-on-solution makes the cluster growth possible in size and sophistication.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    12/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 11

    Advantages of cluster computers

    AvailabilityServer clustering is specifically designed for high availability solution. In case,

    if a server is having a problem another server from the cluster takes over the server

    having issue. This ensures the high availability of the web applications.

    ScalableServer clustering is completely a scalable solution. Resources can be added

    to the cluster accordingly.

    Load handling and maintenanceIf a dedicated server from the cluster requires maintenance, it can be

    stopped while other servers handle its load. Thus, makes the maintenance easier. The

    clustered servers can be configured and managed easily without any trouble.

    Disadvantages of Cluster computing

    More space is requiredServer clustering usually requires more servers and hardware to manage and

    monitor, thus, increases the infrastructure. Some web hosting providers may

    afford it.

    Very expensiveIt is not a cost-effective solution, as it needs a good server design which can be a bit

    expensive.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    13/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 12

    Implementation

    MilitaryThey are used to store the secrete information about various military

    disciplines and can used as storage and backbone for super computers .

    Health careIn the field of health either it is hospital or a pharmaceutical company these

    both will contain the tones of data to managed about their medicines and

    patients so here small clusters can be made.

    EducationIn the field of education whenever we talk about the international level

    exams, lacks of students can simultaneously appear for the same exam

    hence in such situations servers and clusters can play the important role .

    Social networking

    Now it has become a very word for the world millions of users are at a time

    doing social networking there for to store and manage the huge crowd of

    data cluster are used.

    BroadcastingBroadcasting is the one of the famous applications of clusters because tons

    of data are sent and received and simultaneously it is also get stored

    Multimedia(like high definition graphics)Multimedia is closely related to broadcasting because it makes users to

    explore them on the web.

    Very use full in astronomyAstronomy is the very closely related with the clusters one of its example is

    the cluster made by NASA for handling the satellite sent by them.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    14/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 13

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    15/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 14

    Future scope

    etabytes of data can be stored at one place. Petabytes of data or more than it can be accessed quicker. More than one super computer can be combined in single cluster

    that will allow combining a large and vast amount of data to be

    stored and managed from a single area or from a single large

    system.

    In the astronomy more satellites like QURIOSITY (rover).

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    16/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 15

    Conclusions

    The use of clusters of workstations to increase the throughput of hardware andsoftware applications.

    There now exists a significant number of Cluster Management Software (CMS)packages to manage the high workloads and complex situations.

    The importance of cluster software can be seen by both the commercial andtechnical dimensions.

    HTTP protocols could clearly be used as part of CMS packages. Little software of this type has so far been developed at present but several of the

    packages reviewed until but they are under very fast development by experts.

  • 7/22/2019 Cluster Computing by Pritam Bhansali

    17/17

    CLUSTER COMPUTING

    M.Sc-Part-I-CS-Sem-II Page 16

    References:

    1.

    R. Buyya, ed. High Performance Cluster Computing, Volume 1: Architectures andSystems. Prentice-Hall, 1999.

    2. R. Buyya, ed. High Performance Cluster Computing, Volume 2: Programming andApplications, Prentice-Hall, 1999.

    3. J. Salmon, C. Stein, and T.L. Sterling. "Scaling of Beowulf-class DistributedSystems", Proceedings of the Supercomputing '98 Conference, 1998

    4. K. Dowd and C. Severance. High Performance Computing, 2nd ed. O'Reilly andAssociates, 1998

    5. W. Gropp, E. Lusk, and A. Skjellum. Using MPI: Portable Parallel Programming withthe Message Passing Interface. MIT Press, 1994.

    6. C.H. Koebel, D.B. Loveman, R.S. Schreiber, G.L. Steele Jr., and M.E. Zosel. TheHigh Performance Fortran Handbook. MIT Press, 1994.