IBM / UCB EECS Collaboration Meeting May 11, 1999 David E. Culler Computer Science Division U.C. Berkeley.

IBM / UCB EECS Collaboration

MeetingMay 11, 1999

David E. Culler

Computer Science Division

U.C. Berkeley

IBM 5/11/99 intro 2

Meeting Plan

• Where we’ve been– Background on IBM / UCB collaborations (Culler)

• Where we’re going– Universal Computing @ Berkeley (Culler)

– Endeavor Expedition to the 21st Century (Katz)

– Ninja Platform Architecture (Joseph)

» Demo - novel service concepts

– Future Storage Architectures (Patterson)

– Future Digital Libraries (Wilensky)

– Future HCI (Landay)

» Demo - novel interfaces

• How we’re going to get there

• Lunch

• Decisions

IBM 5/11/99 intro 3

IBM/UCB CS Relationship

• < 1992– IBM fellowships were major draw

• 93 - current– relationship structured around SUR grant and executive

liaison

– simple flow of RS6000s & SPs into rest of campus

– CS has broken the mold almost every year

– Successful collaborations

» SUR provided forcing function and opportunity

– Several UCB interns at IBM

• > 1999– How can we amplify each other’s success through major

initiatives?

IBM 5/11/99 intro 4

Brief SUR History

• 1993– SP-1 to Castle (Culler, Demmel, Yelick)

– RS6000s to ME (Bogy), RS6000s to Chem (Miller)

AM, Split-C => 1998 PSSP 2.3 Product - LAPI

Missed Opportunity: NOW

• 1994– RS6000/590s to Castle/NOW (Demmel, Culler, Yelick)

– RS6000s: Lasers (Sheng), RS6000: Comp. Chem (Miller)

ScaLapack => ESSL, PhiPack

• 1995– Tape Robot: Digital Library (Wilensky)

– RS6000s: ME (Bogy)

IBM 5/11/99 intro 5

SUR History (cont)

• 1996– Many raw Disks: Tertiary disk project (Patterson)

» Museum of Fine Arts

– RS6000s: Chem Eng. (Chandler)

• 1997– Disk towers: Millennium (CS, Astro/Phys, Civil, ...)

– G30 SMP: (Yelick, Demmel)

– RS6000s: ME, Chem

• 1998– SP2: Chem (Head-Miller)

– 150 WorkPads: EECS PostPC

– Disk Tower Supplement: Dlib/DataBase Cluster (Wilensky,Hellerstein)

=> Almaden looking at wireless infra.

IBM 5/11/99 intro 6

SUR 1999: Leave the desk behind

• Servers– host production Ninja Services

• Mobile pads– smallest fully networked device to break the tethers

• Emdedded Displays– touch sensitive LCD mounted in the walls

• Cross Pads– extend the NotPals experience

• Vision Pads– change the mode of interaction

IBM 5/11/99 intro 7

Research Collaborations

• Scalable Systems (Culler - Snir, ...)

• Numerical Techniques (Demmel - Gustafson, ...)

• Programming Lang. (Yelick, Aiken - ??)

• Service Support (Joseph - Leiman, ...)

• Mobile Computing (Katz - Krishna)

• Database Theory (Papad. - Raghavan, ...)

• Database (Hellerstein - Carey, Haas, ...)

• Storage Systems (Patterson - ??)

• Digital Libraries (Wilensky - ??)

• Social Networks (Canny - ??)

IBM 5/11/99 intro 8

Universal Computing @ Berkeley

IBM 5/11/99 intro 9

Natural Tides of Innovation

Time

Integration

Innovation

Log R

Mainframe

Minicomputer

Personal ComputerWorkstationServer

2/99

IBM 5/11/99 intro 10

Away from the “average device”

• Powerful, personal capabilities from specialized devices– small, highly mobile or embedded in the environment

• Intelligence + immense storage and processing in the infrastructure

• Everything connected

Laptops, Desktops

Devices

IBM 5/11/99 intro 11

universal

Function: adjective

1 : including or covering all or a whole collectively or distributively without limit or exception

2 a : present or occurring everywhere b : existent or operative everywhere or under all conditions <universal cultural patterns>

3 a : embracing a major part or the greatest portion (as of mankind) <a universal state> <universal practices> b : comprehensively broad and versatile <a universal genius>

4 a : affirming or denying something of all members of a class or of all values of a variable b : denoting every member of a class <a universal term>

5 : adapted or adjustable to meet varied requirements (as of use, shape, or size)

IBM 5/11/99 intro 12

1. covering all without limit

• Internet-Scale Systems Research Group (ISRG)

– MASH / BARWAN / MINT

» Distributed Interactive Services, mobile networks, SRM, NW simulation

– NINJA

» Platform Architecture for Scalable, Customizable Internet Services

– Iceberg

» Computer / Telephony Integration

– Millennium

» Vast Inter-Cluster Testbed

• Digital Libraries

Billions of Apliances

100 Million Wired

Million Scalable Servers

IBM 5/11/99 intro 13

Millennium Testbed

• Large-scale testbed (1000 proc, 12 M$)– Cluster of clusters across campus with high-speed network

– Extended into the workspaces

• Ninja architecture allows developers to “Push Services into the Infrastructure”

Gigabit Ethernet

PDAs Cell PhonesFuture Devices

WirelessInfrastructure

DesktopPCs

Servers

Clusters

Massive Cluster

IBM 5/11/99 intro 14

Core Questions

• How to make it easy to author services that can provide value to a large fraction of the world?

– inktomi serves 20 million users per day!

– Preserve distributed innovation in a service-centric world

– Ninja:

» structured architecture

» push services into the infrastructure

» composition of services

• Design principles of systems at world-wide scale?

IBM 5/11/99 intro 15

2: present or occurring everywhere

• The Ubiquitous, Pervasive connotation

• Small devices + protocols are enablers– everyone is building information appliances

• What is the overall architecture that incorporates these billions of devices?

– How are they programmed?

– Deeply embedded spontaneous networking.

– How do they take on a role in the context of their environment?

– How are they managed?

– How does authentication, security operate?

– Manual override? Crisis?

• How do we interact with a computationally rich environment?

IBM 5/11/99 intro 16

Directions

• Universal Access to Storage as Fundamental Simplifier

– Endeavor Oceanic Store (Katz, Kubiatowitz)

– Massive Storage Systems (Patterson)

• Build services that extend down into devices (Ninja)

– Keep the hardwork in the infrastructure

– Service delivered on a path of proxies

– Construct paths automatically

• Join with TelCo from the start (Iceberg)

• Resource Mgmt via Comp. Economies

• Negotiation Architecture (Endeavor)

• Wildest imagination devices (BSAC)

Wide-Area Path

IBM 5/11/99 intro 17

3. embracing a major part of mankind

• not just building world-wide services

• not just connecting every device

• What fundamentally are the new modes of interaction?

– Novel user interfaces

» very small formats, very large formats, non-WIMPS

– Bridging human differences

– Integrated personal and collaborative environments

– Harnessing Tacit Information

IBM 5/11/99 intro 18

5. adapted or adjustable to meet varied requirements

• Need to write a “program” today that will provide a valuable function to millions of people in execution evironments that don’t yet exist - or haven’t yet been conceived.

• Need fundamentally fluid software– adapts to the structure of its environment

• Any size, any shape

IBM 5/11/99 intro 19

Today’s Plan

• DARPA Endeavor (Katz)– framework for overall effort and industry collaboration

• Ninja / Iceberg / Millennium (Joseph)– novel, scalable services

• Storage Architectures (Patterson)

• Digital Libraries (Wilensky)

• HCI (Landay)

• DISCUSS and DECIDE