Internet2 Distributed Storage Infrastructure Update - PowerPoint PPT Presentation

About This Presentation
Title:

Internet2 Distributed Storage Infrastructure Update

Description:

Geoff Carpenter, German Goldszmidt: Narwhal (IBM) Replication Mechanisms and Modeling ... IBM Research: Narwhal Resolution Proxy. http://dsi.internet2.edu/apps99.html ... – PowerPoint PPT presentation

Number of Views:59
Avg rating:3.0/5.0
Slides: 23
Provided by: Innovative88
Learn more at: https://icl.utk.edu
Category:

less

Transcript and Presenter's Notes

Title: Internet2 Distributed Storage Infrastructure Update


1
Internet2 Distributed Storage Infrastructure
Update
  • Micah Beck
  • Univ. of Tennessee, Knoxville
  • Bert Dempsey
  • Univ. of North Carolina, Chapel Hill
  • Web Caching Workshop BOF
  • 31 March 1999, San Diego
  • http//dsi.internet2.edu

2
I2-DSI Participants
  • UT Knoxville / ICL
  • Micah Beck
  • Terry Moore
  • Martin Swany
  • Judi Talley
  • UNC Chapel Hill /SILS
  • Bert Dempsey
  • Paul Jones (MetaLab)
  • Debra Weiss
  • Zhiwei Xiao
  • GigaPOP and Campus Site Managers
  • UCAID/Internet2
  • Network Storage Working Group
  • Ted HanssApplications Director
  • NC Networking Initiative
  • Digital Library Federation

3
A Word From Our Sponsors
  • Cisco DNS redirection
  • Ellemtel engineering effort
  • IBM large storage DCE servers
  • Novell storage directory servers
  • Starburst reliable multicast software
  • StorageTek large storage servers
  • Sun design collaboration

4
Single Server Model
  • High performance locally
  • Unacceptable performance across commodity backbone

5
Relying on Wide Area QoS
  • High performance access with reserved bandwidth
  • Essential for real-time communication
  • Technically difficult, expensive, not generally
    available

6
I2-DSI Model Replicated Services
  • Clients access nearby server
  • Everyone gets performance
  • Local resources implement a global service

7
I2-DSI Service Architecture
  • Replication
  • Rsynch, Omnicast, AFS/DFSNovell Replication
  • Resolution
  • Sonar DNS, Distributed Director
  • Delegation
  • Cache prefetch

general users
8
Internet Content Channels
  • A channel is a collection of content which can be
    transparently delivered to end user communities
    at a chosen (price,performance) point through a
    flexible, policy-based application of resources

9
Server Channel Examples
  • Replicated Web Servers
  • APIs Standard HTML, Active Server Pages
  • Channels Web sites
  • Streaming Media
  • APIs MPEG-2, proprietary file formats
  • Channels collections of multimedia presentations
  • Executable content
  • APIs Java byte code, Tcl, Perl
  • Channels CGI programs

10
Current Server Deployment
11
IBM Web Cache Manager
RS/6000 AIX Server 1 GB RAM 72 GB Disk / 900 GB
Tape ADSM Heirarchical Storage Mgt.
12
I2-DSI Server Operations
  • Project Operations Coordinator
  • Judi Talley, University of Tennessee at Knoxville
  • Site Managers
  • Dave Vernon, Indiana University
  • David Lassner, University of Hawaii at Manoa
  • Mark Johnson, NC Networking Initiative
  • Michael Rechtenbaugh, EROS Data Center

13
Infrastructure Expansion
  • StorageTek
  • 2 PC/Linux Servers
  • 700GB disk, tape backup (hot!)
  • Novell
  • 6 PC/NetWare Servers
  • 100GB disk
  • Smaller institutions or departments

14
InfrastructureDevelopment Efforts
  • Proximity Resolution
  • Martin Swany SonarDNS
  • Geoff Carpenter, German Goldszmidt Narwhal (IBM)
  • Replication Mechanisms and Modeling
  • Bert Dempsey students
  • Debra Weiss Batch rsync multicast
  • Zhiwei Xiao Network metrics and modeling
  • Channel Representation and Server
  • Leif Abrahamsson, Christophe Achouiantz, Patrik
    Johansson (Ellemtel)

15
I2-DSI Applications Workshop Chapel Hill, NC
March 4 5, 1999
  • 10 applications
  • Indiana Digital music and media library
  • UNC-CH Instructional Management System
  • San Jose State Art history images
  • Vanderbilt zoomable medical images
  • Viagenie Network docs database
  • Columbia Earth sciences environment
  • UNC-CH Virtual Laboratories
  • Ohio Supercomputer Center High Volume Datasets
  • CalTech Globally Interconnected Databases
  • Univ. of Kent National Software Archive
  • Red Hat pan-Linux source distribution

16
I2-DSI Applications Workshop Chapel Hill, NC
March 4 5, 1999
  • 4 technologies
  • Minnesota Scalable Video
  • IBM Research Multicast, Filter and Store
  • Moscow Ctr. for New Info. Tech. in Med. Ed.
    Semantic Text Analysis
  • IBM Research Narwhal Resolution Proxy
  • http//dsi.internet2.edu/apps99.html
  • Special issue of the Journal of Network and
    Computer Applications (Academic Press)

17
Application Management Partner MetaLab.unc.edu
  • The site formerly known as SunSITE.unc.edu
  • Fearless Leader Paul Jones
  • A cool, tall glass of sweet tea on a hot day.
  • 2 M HTTP 1/3 M FTP file transfers daily
  • Collections policy
  • teaching, research, or public service
  • use technology in innovative and unique ways
  • non-commercial or not-for-profit

18
Application Strategy
  • Chose initial applications
  • Available or easily ported services
  • Low update demands
  • Port to an I2-DSI server
  • Our development effort is limited
  • App developers can have access to the servers
  • Distribute to homogeneous core
  • Derive service abstractions

19
The Need for Channel Representation Standards
locally interpreted files
replicated files
Origin Server
Replicated Server
Replicated Server
proxy
Web clients
Standard-based Web traffic
Replication of source files
20
Replication Performance and Scalability Issues
  • Server placement
  • Server resources
  • Server description (metadata)
  • Server Channel description (metadata)
  • Object representation
  • Characterization of replication mechanisms
  • Channel-to-server mapping (subscription)

21
NetStore 99 Workshop
  • Network Storage Technical Workshop
  • Knoxville, TN, October 1999
  • http//dsi.internet2.edu/netstore99
  • Scope
  • I2-DSI implementation
  • I2-DSI applications
  • Related networking projects
  • Storage technology

22
Conclusions
  • A server platform is in place
  • Infrastructure development
  • Service abstractions (search, computation)
  • Publication and replication protocols
  • Portable representation and API
  • Heterogeneous servers
  • Six months to show results from initial
    application development efforts
Write a Comment
User Comments (0)
About PowerShow.com