Grid Computing A summary of concepts and ideas

Grid Computing A summary of concepts and ideas including the implementation there-of. Robert Stallings Thursday, April 20 th, 2006

Contents • Grid Computing – – – – Definition History Reasoning Critical Tasks Infrastructure Architecture Authentication • Globus Toolkit – – – Radford and Globus Installing Globus What’s next Conclusion Questions and Answers

History • Not a new idea – 1965 Fernando Corbato • “Like a power company or water company” – 1968 J. C. R Licklider and Robert W. Taylor

Definition • Server clusters – Clusters of clusters • Grids join multiple server clusters

Why Gird Computing? • Quick Facts – Speed doubles every 12 months – Capacity doubles every 9 months – A computer halves in price every 18 months – Performance of WAN doubles every nine months • Capacity and WAN Performance is comparable

Critical Tasks • Involves using remote resources • Must – Discover they exist – Negotiate access to them – Configure hardware and software – Not lose security.

Critical Tasks • Critical Tasks – Creating and managing services – Single Sign-on – Transferring large datasets quickly – Form large communities – Maintain existence, state, and usage policies

Infrastructure • Internet handles basic communication • Must – Be broadly developed – Simple – Same functionality as the Internet but more • Examples: – – – Science portals Distributed Computing Large scale Data analysis Computer-in-the-loop instrumentation Collaborative work

Architecture

Authentication • More difficult • Computer can be a server AND a client • Must be able to have one computer be able to use resources on another on behalf of the user.

Authentication • Requirements – Single Sign-on – Mapping to local security • Example: Kerberos – Delegation – Community Authorization and Policy

Authentication

Overview Conclusion • Not a new idea • Several Critical tasks • New Infrastructure • New Architecture • Unique Authentication Procedure

The Globus Toolkit Experiences with The Modern Grid

Overview • Radford and Globus • Installing Globus • What’s next • Conclusion • Questions and Answers

Radford and Globus • NSRL – Network and Security Laboratory • CREU – Collaborative Research Environment for Undergraduates • Current Network – 2 Fedora machines – 1 Windows Machine

Globus • The Globus Toolkit – The Globus Alliance • Middleware – GRAM – MDS – GSI – GASS / Grid. FTP

Setting Up Globus • Several Basic Steps – Security – Grid. FTP – Webservices – Reliable File Transfer – WS GRAM

Installation • Installation – Fedora Core 4 • Easy with RPM – Windows • Standard EXE • Linux > Windows – Security – Options – Compatibility

Security • Create user “Globus” • Set up full hostname • Set up dns server • Create a Simple Certificate Authority – Certificates – Certificate Authority – Single Sign-in • Add to Mapfile

Security • 1 st machine as Primary Certificate Authority – Secure – Sign certificates – Distribute • Signing: – Grid-ca-sign –in cert –out signed

Grid. FTP • Start Globus Service – Install xinetd. d • Service Maintainer – Create Grid. FTP Service • Gridftp file • Services file

Grid. FTP • gridftp file service gsiftp { instances socket_type wait user env = 100 = stream = no = root += GLOBUS_LOCATION=/usr/local/globus-4. 0. 1 += LD_LIBRARY_PATH=/usr/local/globus-4. 0. 1/lib 2 server = /usr/local/globus-4. 0. 1/sbin/globus-gridftp-server_args = -i log_on_success += DURATION nice = 10 disable = no }

Grid. FTP • Services File • Add to tail end: # Local services gsiftp 2811/tcp

Webservices • Auto-start container – stop-start script • To start and stop the container easy – Init. d script • To start the stop-start script on bootup • Errors – No Reliable File Transfer (RFT).

Reliable File Transfer • Install Postgres – Database • Create User Postgres • Allow Globus User Admin Privilege

Reliable File Transfer • Create Database – rft_schema. sql • Edit config file with password and user • Restart container

WS GRAM • Use visudo to edit sudoers file – Sudoers • Who can do what (all one line) globus ALL=(fedoralee 1) NOPASSWD: /usr/local/globus-4. 0. 1/libexec/globus-gridmap-and-execute -g /etc/grid-security/grid-mapfile /usr/local/globus-4. 0. 1/libexec/globus-job-manager-script. pl * globus ALL=(fedoralee 1) NOPASSWD: /usr/local/globus-4. 0. 1/libexec/globus-gridmap-and-execute -g /etc/grid-security/grid-mapfile /usr/local/globus-4. 0. 1/libexec/globus-gram-local-proxy-tool * DONE! (with the first machine)

Second Machine • Same procedure except. . – Trusting 1 st machine CA – Certificate Timing – DNS configuration • Copy second machine for duplication

Common Installation Problems • No fully Qualified Domain – Create a full Domain Name Service server • No Mapfile Entry for name • Invalid Timestamps for certificates

Still to Be Done. . • NTP between computers • Testing Security • Porting Globus to First Windows Machines • Better Troubleshooting Guide

Conclusion • A lot of time • Huge learning curve • Globus is powerful

Questions ?