Ten Thousand miles to the DSG - PowerPoint PPT Presentation

1 / 39
About This Presentation
Title:

Ten Thousand miles to the DSG

Description:

Auto Network Topology Discovery Module. Discover the physical topology of any network, ... network as a static entity (No leasing) Infact, why not use SOAP in ... – PowerPoint PPT presentation

Number of Views:50
Avg rating:3.0/5.0
Slides: 40
Provided by: acetR
Category:
Tags: dsg | miles | ten | thousand

less

Transcript and Presenter's Notes

Title: Ten Thousand miles to the DSG


1
Ten Thousand miles to the DSG
  • Aamir Shafi
  • Distributed Systems Group
  • University of Portsmouth
  • Portsmouth, UK
  • Aamir.Shafi_at_port.ac.uk

2
Outline
  • Introduction and why this title ?
  • Previous project experiences
  • DIAMONDS
  • MonALISA
  • Teaching Experience Community Services
  • Suggestions for the DSG
  • Research Interests
  • Middle ware technologies their interfacing
  • Agent-based DS for the Grid
  • Network monitoring and diagnostic tool
  • Evaluation of .NET platform
  • Software Development paradigms
  • Summary

3
Introduction title
  • Aims of the talk
  • Introduce myself and my back ground work
  • What are the research areas I would like to work
    on, being part of DSG
  • Title, well... that's probably very true
  • Belong to Multan
  • Did my BCSEngg from NUST, Islamabad
  • Worked for NUST-Caltech acedemic collaboration
    for one and a half year

4
(No Transcript)
5
Previous Projects and Work experience
6
DIAMOnDS (Distributed Mobile Agents for Mobile
Dynamic Services)
  • Framework for higher level services (Job
    scheduling services) to make use of MonALISA
    monitoring information.
  • Paper published in the proceedings of Computing
    in High Energy Physics (CHEP03), UCSD,
    California, 23-28 March, 03 http//arxiv.org/abs/c
    s.DC/0305062
  • Project Web Site
  • http//www.niit.edu.pk/acadamics/niit_research/ndg
    c/research_ndgc_diamonds.html

7
Why a new Mobile Agent Framework?
  • Web-Based Toolkit
  • Only toolkit that allows agents to be started
    from the web
  • Truly Mobile Agents
  • A requirement to achieve robustness in WAN
    applications. (FIPA-OS moves agent by cloning)
  • Remote Monitoring
  • Only toolkit in which agents have remote GUIs
    that can be downloaded on the client side
  • Integration of SOA
  • The notion of Agents acting as mobile services
    for end user

8
Architecture of DIAMOnDS
9
Architecture of DIAMOnDS
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
MonALISA
  • Goal is
  • Provide monitoring information from large and
    distributed systems to a set of loosely coupled
    "higher level services" in a flexible, self
    describing way.
  • Monitoring what ?
  • Resources which Regional Centers would like to
    share on LHC Data Grid
  • RC's, Farms, abstractions in LHC Data Grid ...
  • Ultimately higher level services would access
    this monitoring data, to make decisions

14
MonALISA (Demo)
  • Project web site http//monalisa.cacr.caltech.edu
  • Iosif Legrand among the main developers
  • MonALISA Demo
  • MonALISA Web Repository

15
(No Transcript)
16
Auto Network Topology Discovery Module
  • Discover the physical topology of any network,
  • First implementation was only in SNMP,currently,
    ping, zonetransfer, nslookup etc are considered.
  • Cornell, takes one hr,CERN, took three hrs
  • Highly CPU intensive module, parallel
    implementation may be interesting
  • Details -http//www.niit.edu.pk/acadamics/niit_re
    search/ndgc/resources_presentations/network_topolo
    gy_discovery.ppt

17
Little Monalisa
  • What about monitoring the clients, who make use
    of Grid, Web, Jini services ?
  • Little MonALISA, part of Internet2 End to End
    Performance Initiative work
  • Demo
  • http//monalisa.cern.ch/LML1/lml.jnlp
  • Web Site http//202.125.153.148/newlml5/index.h
    tml
  • Future work, aiming to build a diagnostic tool

18
TCP Monitor and Auto Tuner
  • Web100 project, http//www.web100.org
  • Implementation of linux kernel, and C library
    that allows to extract a lot of info about TCP
    connections.
  • Playaround with C/R/S Window Size, SACK,
  • Build a GUI that presented TCP connection
    information
  • Around 100 different TCP parameters, (adjustable)
  • Java Analysis Studio (JAS) enabled GUI

19
Teaching Experience Community Services
  • Course included
  • .Net Introduction
  • Syntax of C
  • ASP.Net ADO.Net
  • My part to teach, Jini and Apache Axis
  • VRVS support person (http//www.vrvs.org)
  • Managing XEON dual processor running MonALISA,
    VRVS Reflector
  • Managing Windows 2000 Domain Server

20
Suggestions for DSG
  • Java Web Start
  • Kick start any Java application from a web page
  • Power of Java2D, Java3D, Swing at your disposal
  • Impressive for Demos (sample in next slide)
  • ANT (build tool)
  • Platform independent
  • Compile, Clean, MakeJars, and even execute by one
    command
  • Supported by Sun
  • VRVS support center for UK
  • Installating MonALISA

21
Sample JNLP (Java Network Lauching Protocol) File
lt?xml version"1.0" encoding"utf-8"?gt ltjnlp
spec"1.0" codebase"http//dsg.port.ac.uk/shafi
a" href"aamir.jnlp"gt ltinformationgt
lttitlegtMPJ Clientlt/titlegt
ltvendorgtDSGlt/vendorgt lthomepage
href"http//dsg.port.ac.uk"/gt
ltdescriptiongtExample file for
dsglt/descriptiongt lt/informationgt
ltsecuritygt ltall-permissions/gt
lt/securitygt ltresourcesgt
ltj2se version"1.4"/gt ltjar
href"Aamir.jar"/gt ltjar
href"Shafi.jar"/gt ltjar
href"tools.jar"/gt lt/resourcesgt ltapplication-de
sc main-class"MainClass"/gt lt/jnlpgt
22
Sample build.xml file
ltproject name"MyProject" default"dist"
basedir"."gt ltdescriptiongt simple
example build file lt/descriptiongt lt!-- set
global properties for this build --gt ltproperty
name"src" location"src"/gt ltproperty
name"build" location"build"/gt ltproperty
name"dist" location"dist"/gt lttarget
name"init"gt lt!-- Create the time stamp --gt
lttstamp/gt lt!-- Create the build directory
structure used by compile --gt ltmkdir
dir"build"/gt lt/targetgt lttarget
name"compile" depends"init"
description"compile the source " gt lt!--
Compile the java code from src into build
--gt ltjavac srcdir"src" destdir"build"/
gt lt/targetgt
23
Sample build.xml contd ....
lttarget name"dist" depends"compile"
description"generate the distribution" gt
lt!-- Create the distribution directory --gt
ltmkdir dir"dist/lib"/gt lt!-- Put
everything in build into the
MyProject-DSTAMP.jar file --gt ltjar
jarfile"dist/lib/MyProject-DSTAMP.jar"
basedir"build"/gt lt/targetgt lttarget
name"clean" description"clean up" gt
lt!-- Delete the build and dist directory
trees --gt ltdelete dir"build"/gt
ltdelete dir"dist"/gt lt/targetgt lt/projectgt
24
Research Interests
1. Performance and Feature Evaluation of Middle
ware Technologies
25
Jini
  • Amazing features
  • Leasing
  • Code Mobility
  • Distributed Events Model
  • Transaction support
  • JavaSpace (interesting feature to provide fault
    tolerance)
  • Protocol Independence (amazing ability, can be
    used for interfacing)
  • Excellent security support
  • Surrogate Architecture
  • Not so amazing things
  • Usage of RMI
  • Maybe JERI helps ?
  • Java dependence

26
Web Services
  • Amazing Features
  • Platform and Language independence
  • Use of XML (SOAP)
  • Not so amazing features
  • Security
  • Transaction
  • Treats network as a static entity (No leasing)
  • Infact, why not use SOAP in Jini instead of RMI ?

27
Other Technologies
  • Globus (GT3) (http//www.globus.org)
  • JXTA (http//www.jxta.org)
  • GigaSpaces (http//www.gigaspaces.com)
  • Commercial implementation of JavaSpaces
  • Software infrastructure for developing clustered,
    synchronized and service oriented applications
  • Acts as a real time data bus, using distributed
    shared cache (Gigaspace)
  • Is interesting and research license for one year
    can be obtained

28
Interfacing
  • Judy Project, (phases)
  • Jini client to access a web service
  • Jini service to be advertised as Web service
  • LUSs act like UDDI database
  • UDDI database act like LUS
  • Being a non-active developer of this project, I
    plan to work for this.
  • http//www.judy.jini.org, Owner R Dale Asberry
    and till date, the sole programmer as well.
  • Embryonic stage, but once implemented will be a
    proof of concept

29
Interfacing other technologies
  • Using JXTA protocols to improve Jini discovery
    protocols on WAN (as identified in Jini meets the
    Grid
  • OGSI interfacing with Jini
  • OGSI Java programming API
  • GigaSpaces is already developed in Java/Jini
  • Once done, Grid Jini

30
Research Interests
2. Investigation of agents-based Distributed
Systems in Grid
31
Agent-Based DS for Grid
  • Many toolkits available
  • IBM Aglets, FIPA-OS, JADE (very popular)
  • Do they really ?
  • Reduce network load and overcome latency
  • Execute asynchronously and autonomously.
  • Naturally heterogeneous, adapt dynamically
  • Are fault-tolerant and allow disconnected
    operations
  • If yes, then we should adopt agents for message
    passing in DS
  • Maybe use mobile agents for message passing in
    MPJ (if on WAN)

32
Research Interests
4. Performance and Feature comparison of C and
Java, infact, the .NET platform and Java
33
.Net and Java (J2EE)
  • All know, C resembles Java a lot
  • My experience (naïve) suggests,
  • Excellent platform for RAD
  • ASP.NET, much powerful than other scripting
    languages
  • Why doesnt any researcher in DS uses it ??
  • A lot of papers already written
  • Evaluate .Net for HPC
  • Compare its performance with Java

34
Research Interests
5. Software Development paradigm
35
Why OOSE ??
  • Only 10 of software projects are in-time and
    in-budget.
  • Studied two courses
  • Learnt DFD(Data Flow Diagrams), ERD (entity
    relationship diagrams), UML, Rational Rose, Objec
    cts, Procedures .
  • In short I wasted my time, each of the design
    model and tool is a failure
  • Extreme Programming (XP)
  • Aspect Orient Programming
  • Would like to make my own software designing tool
    until I find one .

36
Research Interests
5. Building a network monitoring diagnostic
tools
37
  • Not so long ago, Quality of VRVS
  • Caltech --gtgt Excellent
  • CERN --gtgt Very Poor
  • E mailed Internet2(http//www.internet2.edu)
    network experts, solution was
  • Run traceroute to determine the path
  • NUST-Caltech
  • NUST-CERN
  • Run Iperf with CERN and Caltech
  • Get back ground traffic
  • What through put TCP can achieve

38
  • Email from the experts continue ...
  • Test at different times of the day to determine
    if congestion out of Pakistan is the culprit
  • If Congestion, then apply ltltDivide and Conquergtgt
    between sites
  • Cant even tell, why my voice chat is not working.
  • Not one tool to pin point, diagnose the problem
  • Noticeable in slower networks
  • Would like to explore this area,
  • Web100 may help

39
Summary
  • Introduced my self and my previous work
  • Discussed my research interests
  • Looking forward to contribute, while being part
    of DSG

?
Write a Comment
User Comments (0)
About PowerShow.com