Scalable System Design

Embed Size (px)

Citation preview

  • 8/19/2019 Scalable System Design

    1/22

     A presentationon

    Scalable System Design1

    Presented by:

    Md. Hasan Ansar i

    MSCS/071/655

  • 8/19/2019 Scalable System Design

    2/22

    Inroduction

     Scalability  is ability of a system, network, or process tohandle a growing amount of work in a capable manner.

    For example, it can refer to the capability of a system toincrease its total output under an increased load whenresources (typically hardware) are added.

    It is a highly significant issue in electronics systems,

    databases, routers, and networking. A system, whoseperformance improes after adding hardware,proportionally to the capacity added, is said to be ascalable system.

    !

  • 8/19/2019 Scalable System Design

    3/22

    "easuring the scalability of a system#

    $here hae also been different perceptions on how tomeasure scalability% maximum number of simultaneous users supported by

    the system or maximum transactions processed per unit of time by the

    system.

    For example, the re&uirement could say that the systeminitially should be able to support ' simultaneous users, but it should be scalable enough to support up to 'simultaneous users.

  • 8/19/2019 Scalable System Design

    4/22

    "easuring the scalability of a system

    • $he throughput of thesystem usually iincreasesinitially as the system isexposed to an increasing

    load.

    •  After a certain limit thethroughput of the system

    decreases from *neepoint.

  • 8/19/2019 Scalable System Design

    5/22

    "easuring the scalability of a system'

    $his may happen due tomany reasons

    like the system may befeeling the scarcity ofhardware resources

     After adding hardwareknee point is shifted.

  • 8/19/2019 Scalable System Design

    6/22

    +calable +ystem esign -atterns

    /ptimi0e Algorithm

     Add ardware

    Introduce -arallelism

    /ptimi0e ecentrali0ation2ontrol +hared 3esources

  • 8/19/2019 Scalable System Design

    7/22

    /ptimi0e Algorithm

    4

    $he key to the solution is to identify the areas those can beoptimi0ed for performance when the input load increases.

    $he aim is to identify tasks which can be completed in ashorter period to sae processing time. $his shall result inoerall throughput improement of the system by allowingthe saed 2-5 time to be allocated for growing work.

    $hese areas can in some cases be identified by carrying acode walkthrough to identify areas where smarteralgorithms can help improing the performance.

  • 8/19/2019 Scalable System Design

    8/22

    /ptimi0e Algorithm

    6

    If code walkthrough does not help, then certain tools can be usedto identify the areas which are consuming most of the timeduring processing.

    $hese tools (like I7" 3ational 8uantifier) help identifying theareas (functions) which are eating up most of the 2-5 time.

    5sing an alternate algorithm which shall result in keeping theend result same but complete the same task in a shorter duration.

    *nown 5ses% 3unning time of Insertion sort is / (n9n), whilethat of 8uick sort is /(n lg n). ence, &uick sort scales better with n as compared to insertion sort as depicted in the Figure .

  • 8/19/2019 Scalable System Design

    9/22

    /ptimi0e Algorithm:

  • 8/19/2019 Scalable System Design

    10/22

     Add ardware1

    $he key to the solution is to identify the hardware resources which are becoming scarce for the system. $he scarce hardwareresources can be identified by using resource monitoring tools(like prstat;top).

    /nce the scarce hardware resources are identified, the next stepis to add them in efficient &uantity. $he hardware resources can be added to the same existing physical hardware or it may beadded as a new physical hardware.

     Adding hardware to the existing physical node will result in whatis commonly known as

  • 8/19/2019 Scalable System Design

    11/22

     Add ardware11

    *nown 5ses%

     Adding 3A" to the existing machine incase it isdetected that the system re&uires more memory.

     Adding a separate machine to a cluster in case it isfound that all the existing machines in the cluster are

     being utili0ed to their full capacity.

  • 8/19/2019 Scalable System Design

    12/22

    Introduce -arallelism1!

    ow can a system maintain its performance withincreasing input load>

    +ystem that does not want to deteriorate itsperformance when the load increases and has thecapability to split the work in to pieces which can be completed simultaneously.

  • 8/19/2019 Scalable System Design

    13/22

    Introduce -arallelism1#

    $he key to a scalable design is to process multipletransactions in the system simultaneously. $hesystem should do parallel processing in order tomaintain the throughput with the increasing load.

    $he work should be diided in to pieces that can bedone simultaneously so as to do more processing in

    the same time.

  • 8/19/2019 Scalable System Design

    14/22

    Introduce -arallelism1

  • 8/19/2019 Scalable System Design

    15/22

    Introduce -arallelism1'

    -arallelism can come in different forms%

    ? A system can hae multiple threads (Intra?processscalability)

    ? /r, it can hae multiple processes (Inter?processscalability)

    ? /r, a system can hae mix of both of the aboe(ybrid scalability) in order to process multipletransactions simultaneously.

  • 8/19/2019 Scalable System Design

    16/22

    Introduce -arallelism1

    *nown 5ses%

     Apache?$omcat cluster haing multiple $omcatnodes @ each one of them processing

    $$- re&uests independently. oweer, thesedifferent instances may be talking to the samedatabase instance.

  • 8/19/2019 Scalable System Design

    17/22

    7roadband 3emote Access +erer

      A 2ase +tudy in epal $elecom14

     A broadband remote access server (BRAS) routestraffic to and from broadband remote access deicessuch as digital subscriber line accessmultiplexers (+BA") on an Internet serice

    proiderCs (I+-) network. 73A+ can also be referred toas a Broadband Network Gateway  (BNG).

    $he 73A+ sits at the core of an I+-Cs network, andaggregates user sessions from the access network. It isat the 73A+ that an I+- can inDect policy managementand I- 8uality of +erice (8o+).

  • 8/19/2019 Scalable System Design

    18/22

    7roadband 3emote Access +erer

    16

  • 8/19/2019 Scalable System Design

    19/22

    7roadband 3emote Access +erer1:

    $he specific tasks include% Aggregates the circuits from one or more link access deices

    such as +BA"s

    -roides layer ! connectiity through either transparent bridging or --- sessions oer Ethernet or A$" sessions

    Enforces &uality of serice (8o+) policies

    -roides layer # connectiity and routes I- traffic throughan Internet serice proiders backbone network to theInternet

  • 8/19/2019 Scalable System Design

    20/22

    7roadband remote access serer

    !

     A +BA" collects data traffic from multiple subscribers into acentrali0ed point so that it can be transported to a switch orrouter oer a Frame 3elay, A$", or Ethernet connection.

    $he router proides the logical network termination. 2ommonlink access methods include --- oer Ethernet (---oE), ---oer A$" (---oA, bridged ethernet oer A$".

    $he 73A+ is responsible for assigning network parameters suchas I- addresses to the clients.

    $he 73A+ is also the interface to authentication, authori0ationand accounting systems.

  • 8/19/2019 Scalable System Design

    21/22

    3eferences

    !1

  • 8/19/2019 Scalable System Design

    22/22

    !!

    $hanks

    8ueries>