Jim Dim Talk at College of Tokyo.


111 views
Uploaded on:
Category: Animals / Pets
Description
Jim Dim Talk at College of Tokyo Individual perspectives on PITAC report: invest in long haul research Review of Turing address: 10 long haul research issues Shrubbery: Compress information in the internet Turing: Astute PCs 7 9s: manufacture frameworks that are dependably up and demonstrate it. 5-Moment principle
Transcripts
Slide 1

Jim Gray Talk at University of Tokyo Personal perspectives on PITAC report: invest in long haul exploration Preview of Turing address: 10 long haul research issues Bush: Summarize information in the internet Turing: Intelligent Computers 7 9s: manufacture frameworks that are dependably up and demonstrate it. 5-Minute standard For plates For tapes Sorting Progress PennySort Terabyte Sort (!) Slides will be at http://research.Microsoft.com/~Gray/talks

Slide 2

Presidential Advisory Committee on High Performance Computing and Communications, Information Technologies, and the Next Generation Internet Information Technology http://www.ccic.gov/air conditioning/interval/or http://research.microsoft.com/~Gray/papers/PITAC_Interim_Report_8_98.doc

Slide 3

Charter for the Committee: give an autonomous appraisal of High-Performance Computing and Communications (HPCC) Progress Balance among examination segments; Next Generation Internet activity; Progress Balance IT Research and improvement Maintain United States authority in IT and Applications

Slide 4

Committee Members Co-Chairs: Bill Joy, Sun Microsystems • Ken Kennedy, Rice University Members: Eric Benhamou, 3Com • Vinton Cerf, MCI Ching-chih Chen, Simmons • David Cooper, LLNL Steve Dorfman, Hughes • David Dorman, PointCast Bob Ewald, SGI • David Farber, U. of Pennsylvania Sherri Fuller, U. of Washington • Hector Garcia-Molina, Stanford Susan Graham, UC Berkeley • Jim Gray, Microsoft Danny Hillis, Disney, Inc • John Miller, Montana State Univ. David Nagel, AT&T • Raj Reddy, Carnegie Mellon Ted Shortliffe, Stanford • Larry Smarr, U. of Illinois @ UC Joe Thompson, Miss. State U. • Les Vadasz, Intel Andy Viterbi, Qualcom • Steve Wallach, Centerpoint Irving Wladawsky-Berger, IBM

Slide 5

My Summary of the Report 1/3 of the US financial development since 1992 was in the IT segment. IT is critical to our wellbeing, riches, and security. Made 400 B$ of riches in most recent 3 years (!!) Federal IT exploration subsidizing of a quarter century, made the blast. Government IT examination financing for the most recent decade has been level (in consistent dollars). Exploration subsidizing is progressively close term & connected improvement The board of trustees prescribes Increase long haul examination financing in: Software configuration and usage advancements Technologies to scale the Next Generation Internet to 6 billion clients. Devices, calculations, and frameworks for superior figuring. Spend a billion dollars throughout the following 5 years on Lewis and Clark style "expeditions" into the internet.

Slide 6

Myths Now that IT is a major business, Industry will do long haul research. Actuality: industry spends LITTLE on long haul research. it is not to their greatest advantage IT research = purchase PCs for researchers. Actuality software engineering examination is unique in relation to the use of PCs to some order.

Slide 7

Research Priorities Findings: Total elected Information innovation R&D speculation is lacking Federal IT R&D is too much centered around close term issues Recommendations: Create a key activity in long haul IT R&D Increase the venture for examination in software, scalable data framework, high-end registering, and socio-monetary and workforce sways

Slide 8

Software Research Findings: Demand for programming far surpasses the nation’s capacity to deliver it The country relies on upon delicate programming Technologies to manufacture solid and secure programming are insufficient The country is under-putting resources into central programming exploration Recommendations: Fund more crucial examination in programming advancement techniques and segment advances Sponsor a national library of programming segments Make programming research a substantive segment of each real IT research activity Support research in human-PC interfaces and association Make basic programming research an outright need

Slide 9

Scalable Information Infrastructure Findings: The Internet has developed well past the aim of its unique fashioners Our nation’s reliance on the data base is expanding day by day We can\'t securely stretch out what we as of now know not complex frameworks Learning how to assemble extensive scale, exceedingly dependable and secure frameworks obliges research Recommendations: Increase subsidizing in innovative work of center programming and correspondences advances pointed straightforwardly at the test of scaling the data base Expand the Next Generation Internet test beds to incorporate extra industry organizations keeping in mind the end goal to cultivate the quick commercialization and arrangement of empowering advances

Slide 10

High-End Computing Findings HEC is: fundamental for science and building research a component of the United States national security ready for new applications suppliers experience the ill effects of abnormal business sector weights Research& Development Recommendations Fund imaginative advances and architectures Fund HEC programming (parallel programming) Aim for a genuine application petaops by 2010 through a both equipment and programming procedures Fund HEC frameworks for science and designing examination

Slide 11

Social, Economic, Workforce Recommendations Expand research on the social and financial effects of data innovation dispersion and appropriation Expand activities to expand IT proficiency, get to and research abilities Address the deficiency of high-innovation specialists Programs to re-train “stale” IT laborers Encourage cooperation by ladies and minorities Short-term increment in movement of gifted IT specialists

Slide 12

Conclusions IT is a vital establishment for trade, instruction, medicinal services, ecological stewardship, and national security: Dramatically change the way we impart, learn, manage data and behavior examination Transform the way of work, nature of business, item plan cycle, routine of health awareness, and the administration itself The aggregate Federal IT R&D venture is deficient The Federal IT R&D is unnecessarily centered around close term issues U. S. government must: Create a vital activity in long haul IT R&D Establish a successful structure for overseeing and organizing IT

Slide 13

Jim Gray Talk at University of Tokyo Personal perspectives on PITAC report: invest in long haul examination Preview of Turing address: 10 long haul research issues Bush: Summarize data in the internet Turing: Intelligent Computers 7 9s: fabricate frameworks that are dependably up and demonstrate it. 5-Minute tenet For plates For tapes Sorting Progress PennySort Terabyte Sort (!) Slides will be at http://research.Microsoft.com/~Gray/talks

Slide 14

Vanaveer Bush: Memex: Proposed putting all data online (1948) It will happen Result: InfoGlut. A lot of data in the shoebox Challenge : Organize the data. Give answers comparable to a specialist in the field. Expect inquiries thus illuminate “subscriber” Protect individual security A programmer can\'t become acquainted with your own data without your assent.

Slide 15

Turing’s Test (1951): Intelligent Machines Computers assisted with the 4-shading issue end amusement Computers (and individuals) won world chess title Computers will probably be our 5 th mind Augment our knowledge See for us, hear for us, read for us, Prosthetic eyes, ears, voices, arms, legs,…. Presumably PCs will be wise like plants and creatures. Maybe PCs can be canny like individuals Pass the Turing Test (simple/unthinkable?) (70%, 5 minutes, B can lie) Translating phone (tantamount to a human interpreter) Read a course book and pass the composed exam. Pass a graduate programming class Pass a graduate writing class Radical: Download somebody.

Slide 16

Dependable Systems Build a framework utilized by a large number of individuals every day. At that point: Prove that it does what it should do (code matches spec). Demonstrate that it conveys 99.99999% (7 9s) accessibility (1 hr for every thousand years) Prove that it can\'t be “hacked” for under 1B$ (Y2K $) Then form the framework consequently from the particular.

Slide 17

Jim Gray Talk at University of Tokyo Personal perspectives on PITAC report: invest in long haul exploration Preview of Turing address: 10 long haul research issues Bush: Summarize information in the internet Turing: Intelligent Computers 7 9s: fabricate frameworks that are dependably up and demonstrate it. 5-Minute standard For circles For tapes Sorting Progress PennySort Terabyte Sort (!) Slides will be at http://research.Microsoft.com/~Gray/talks

Slide 18

Storage Hierarchy (9 levels) Cache 1, 2 Main (1, 2, 3 if nUMA). Plate (1 (stored), 2) Tape (1 (mounted), 2)

Slide 19

Meta-Message: Technology Ratios Are Important If everything gets speedier & less expensive at the same rate THEN nothing truly changes. Things showing signs of improvement: correspondence speed & cost 1,000x processor speed & cost 100x capacity size & cost 100x Things staying about the same pace of light (pretty much consistent) individuals (10x more extravagant) stockpiling rate (just 10x better)

Slide 20

15 4 10 12 2 10 9 0 10 6 - 2 10 3 - 4 10 Today’s Storage Hierarchy : Speed & Capacity versus Cost Tradeoffs Size versus Speed Price versus Speed Cache Nearline Tape Offline Main Tape Disk Secondary Online $/MB Secondary Tape Disk Typical System (bytes) Main Offline Nearline Tape Cache - 9 - 6 - 3 0 3 - 9 - 6 - 3 0 3 10 Access Time (seconds) Access Time (seconds)

Slide 21

Storage Ratios Changed 10x better get to time 10x more data transfer capacity 4,000x lower media value DRAM/DISK 100:1 to 10:10 to 50:1

Slide 22

M$: expense of a RAM page RAM $/MB PageSize x Lifetime A$: expense of a circle access Disk Price AccessesPerSec x Lifetime RI: Reference Interval time between gets to

Recommended
View more...