Grid Computing

1. Grid Computing Its Promise and Challenges Tom Smith Master’s Candidate Computer Science Union College January 2004

2. What is Grid Computing? • The word “grid” is chosen by analogy with the power grid: • “always on” resource • Users don’t need to know where the resource is coming from • Global (or at least national) standard • The Grid may be as revolutionary as the power grid and the PC were at their time of introduction.

3. What is Grid computing? • Classic example: SETI@Home • The basic definition (for today): Grid computing is assembling more than machine or system into a unified resource. • Little agreement about what exactly constitutes a “grid” • Refer to “progression of parallel computing”

4. What is Grid computing? • Ian Foster: Luminous figure in the field, based at U of Chicago and Argonne National Laboratory • Argues for a formal definition of the grid

5. Foster’s Definition • Grids are motivated by virtual organizations • Grids are heterogeneous in many ways: • Diverse machine configurations • Diverse network issues • Diverse authentication methods • “Real” grids embrace heterogeneity

6. Foster’s Definition, cont’d • Grids should be able to talk to one another (intergrid protocols) • Most important: grids allow access to a variety of resources (not just cycles) • Sensors • Storage • Databases • Memory • Applications • Etc.

7. Early Success: SETI@Home • Search for ExtraTerrestrial Intelligence • Runs as a screensaver/background app • 35GB daily tapes split into 250kb chunks • Almost 5 million users • ~1500 signed up yesterday • 1,781,106.179 years of CPU time has been contributed to date

8. Early Success: United Devices • Based on intranet at Novartis (drug research) • Molecular modeling • Alternative to buying a new high-performance computer • Saved $2,000,000 with a $400,000 investment

9. United Devices: cont’d • Projects that would have taken 6 years, now take 12 hours. • Initial test was on 2,700 corporate PCs • So successful, installing it on all 70,000 corporate desktops

10. United Devices: Also available • Grid MP Alliance: sell your corporation’s excess cycles to UD • Grid MP Global: “rent” cycles from UD • Implementation of something that seemed only theoretical short time ago • “Computing on demand”

11. Challenges of Grid computing • Security • Bandwidth/networking • Authentication/authorization • Storage • Reliability: what if a node goes down? • Reporting • Synchronization • Nearly every aspect of computer architecture touched.

12. Strategies for Simplifying Problem • Virtualization/abstraction • Virtual OS • Standards

13. Anatomy of the Grid • Influential paper, 2001: Foster et al. • Introduced notion of virtual organization • Suggested “an open and extensible grid architecture”

14. Physiology of the Grid • Another influential paper • Foster et al., June 2002 • New concepts: • Open Grid Services Architecture (OGSA) • Shift in focus from “resources” to “grid services” • Grid services are a special instance of web services

15. Web Services • Cross-platform way of letting different systems interact and exchange information • Based on WSDL and SOAP

16. SOAP • Simple Object Access Protocol • Maintained by W3C, current ver. 1.2 • XML-based, runs on top of HTTP • Simple, platform-independent way of exchanging info between 2 endpoints • An “envelope” protocol • Click here for example

17. WSDL • Web Services Description Language • Developed/Supported by IBM, Microsoft • Proposed standard to W3C, latest proposal is version 1.2 • XML-based, works in conjunction with SOAP, HTTP, MIME • portType: defines function library or class that is being used/requested • Click here for example • Click here for Complete WSDL example w/SOAP binding

18. OGSA and OGSI • OGSA = Open Grid Services Architecture • Web Services + standard grid interfaces = grid services • OGSI = Open Grid Services Infrastructure • OGSI v1.0 was released by Global Grid Forum in June 2003 • Click here for illustration of OGSA • Click here for illustration of OGSI

19. Note: Globus Toolkit • Open-source collection of libraries and functions for implementing grids • “De facto standard” for grid software • Foster involved from the beginning • Generated much of OGSA, OGSI

20. Adoption of OGSA/OGSI • Initially, confusion reigned (January 2003) • Situation clearer after release of OGSI (June 2003) • Most of the grid vendors pay at least lip service to OGSA/OGSI • WSDL’s full incorporation of grid concepts may make OGSA/OGSI less important • For example: United Devices’ software makes use of SOAP and WSDL

21. “The” Grid • Akin to “The internet” • Still just a concept • No grids talking to each other in any significant ways yet • Lots of obstacles to overcome first

22. Case Study: GridFTP • GridFTP was a protocol developed by Foster et al. for Globus Toolkit 2. • Click here for image.

23. Motivations for Grid Computing • High-performance computing • Processor cycle recovery • Resource sharing • Collaboration • Approaches have been parallel so far, but are converging.

24. Questions?

25. Question • Gregory Andrews: “Even most personal computers will soon have a few processors” (2000) • Will the grid make multiprocessor computers irrelevant?

26. Question • Will “The Grid” become as pervasive as the internet?

27. Question • What about bandwidth issues? • How will the irregular rollout of high-speed networks affect the expansion of the grid?

28. Question • Will home PCs play any significant role in high-performance computing? • Will organizations without the need for high-performance computing rent out their resources? • How will the grid change the nature of the PC?

29. Question • According to The New York Times, Europe is exceeding the U.S. in their implementation of grid technologies. • Is this important?