SimMillennium Project Overview David E. Culler Computer Science Division U.C. Berkeley NSF Site Visit March 2, 1998

SimMillennium Project Overview

David E. Culler

Computer Science Division

U.C. Berkeley

NSF Site Visit

March 2, 1998

March 2, 1998 SimMillennium Overview 2

The Vision

• To work, think, and study in a computationally rich environment with deep information stores and powerful services– test ideas through simulation

– explore and investigate data and information

– share, manipulate, and interact through natural actions

• Organized in a manner consistent with the University setting


SimMillennium Project Goals

• Enable major advances in Computational Science and Engineering– Simulation, Modeling, and Information Processing becoming

ubiquitous

• Explore novel design techniques for large, complex systems– Fundamental Computer Science problems ahead are problems

of scale

• Develop fundamentally better ways of assimilating and interacting with large volumes of information– and with each other

• Explore emerging technologies– networking, OS, devices


Goals of this talk

• Components of the Project– Community

– Cluster-based Resouces

– Connectivity

– User Interaction

– Computational Economics

• Specific Infrastructure

• Research Agenda


Component 0: Community

• An inter-disciplinary community with common interests and shared view of the future– strong momentum in computational science and engineering

=> Session II

– Members of 17 campus units and NERSC in Intel Millennium

– Need and commitment required for participation

– Key subset represented in this proposal


Component 1: Resources (Millennium)

• An environment with vast cluster-based computing power and storage (CLUMPS)

behind a personal 3D desktop

NT 3DDesktop

GroupClusterof SMPs

Dept.SMP

CampusCluster


Resource Component Support

• Computers via Intel Technology 2000 grant– 200 NT desktops

– 16 department 4-way SMPs

– 8 5x4 Group Clusters,

– 1 ~100x4 Campus Cluster

– PPro => Pentium II => Merced

• Additional storage via IBM SUR grant– 0.5 TB this year => 4 TB

• NT tools via Microsoft grant

• Solaris x86 tools via SMCC grant

• Campus provides Technical staff

• Research provides the prog. and system support

200 Gflop/s

150 GB memory

8 TB disk


Key NSF investment: Cluster Network• Transforms large collection of individual resources into a powerful system

– can be focus on a problem

• High Bandwidth– scales with the number of processors (Gb/s per proc)

• Low Latency

• Low Overhead

• Low Cost

• Simple and Flexible

• Almost no errors

• Low Risk

• Today: Myrinet


Cluster Research Agenda

• Applications grow into resources– huge range of needs

– require Algorithmic Innovation, Prog. Tools, & Performance

• Dealing Deep Memory Hierarchy – New numerical algorithms on CLUMPs

– New compiler techniques for parallel object language

• Fast Multi-protocol Communication

• Global system at large scale– Unix vs. NT, single system image vs. objects

• Exciting technology turnover– VIA, SANs, Gigabit Ethernet

=> Session III


Campus Support Enables Research

• Technical staff develop and deploy common solution and environment– networked systems designer

– unix-based programmer and cluster system admin.

– NT-based programmer and tools

• Technical computing software developed jointly with NERSC

• Participating departments provide system administration and construction costs.


Component 2: Connectivity

• Create a richly interconnected pool of resources owned by members of the community– Enable transportation of huge data sets and computation

– Enable remote visualization and collaboration

– Enable extensive sharing of resources

• Expand networking technology

Campus Cluster

CS Cluster

EECluster

CE Cluster

ME Cluster

Astro/Phys Cluster

xport Cluster

BIO Cluster

Econ/Math Cluster


NSF Investment: Inter-cluster network

• Gigabit Ethernet connecting group clusters and campus cluster

• Bay Networks provides 70% discount

• Campus provides fiber plant, maintenance, and staff


Physical Connectivity


Inter-Cluster Research Agenda

• Vastly expands the scope of systems challenge– integrate well-connected resources according application

needs, rather than physical packaging

– resource allocation, management, and administration

• Network bandwidth matches display BW– Protocols and run-time sys. for visualization, media transport,

interaction, and collaboration.

• Community can share non-trivial resources while preserving sense of ownership– Bandwidth translates into efficiency of exchange

– Data can be anywhere

• Important networking technology in its own right.– Layer 3 switching, QoS, VLan

=> Session III, V


Component 3: User Interaction

• High-quality 3D graphics emerging on cost-effective platforms– desktops and dedicated cluster nodes

– NERSC team provides modern scientific visualization support

• Gigabit network allows this to be remote.

• New displays create “workbench” environment where large volumes of information can be viewed and manipulated.

• Trackers and Haptic interfaces greatly enhance degrees of user input– 3D capture


NSF Investment: UI Technology

• Two Projection Table– large field of view in horizontal (or vertical) orientation

• Phantom Haptic Interface– 3D force feedback

• Motion Tracker– untethered position

• 3D Shutter Glasses– low cost visualization


User Interaction Research Agenda

• Expand access to 3D visualization– Explore any data anywhere

– Ease development

• Develop lab-bench metaphor for Viz– two hands, physical icons

• Fast prototyping and exchange through Informal Interfaces– sketching

• Dealing with large volumes of information– lenses, brushing and linking

• 3D collaboration and interaction

=> Session VI


Component 4: Computational Economy

• How is this vast, integrated pool of resources managed?

• Traditional system approach: empower global OS to provide “optimal” allocation to blind applications– predefined metric, tuned to fixed workload

– ignores the inherent adaptation of demand

• Computer Center– charge => director-to-user feedback according to cost

• Economic view: decentralized allocation according to perceived value– pricing => user-to-user feedback

– compatible niches,sense of control, cooperation

– idea has been around, why now?


Research Agenda

• Natural fit to academic structure– members want control over own resources, and each has

varying needs that far exceed dedicated resources

– incentive for maintaining resources up to par

• Address partial or delayed information, component failure, and user satisfaction from the start

• Framework for elevating design from resources to services

• Rich body of theory, little empirical validation– experts in several parts of the community

• New paradigm for algorithms & perf. Analysis

• Complex, large-scale systems


Basic Approach• Desktop an active agent conducting automated negotiation for resources• Servers provide resources to highest bidders

– monitor usage and enforce limits within remote execution environment– placement based on economic advantage

• Higher level system functions are self-supporting– resource availability, brokering, directories

• Useful applications packaged as services– may charge more than resources cost


NSF Investment: Staff Support

• Provide enabling technology and let it evolve– monitoring, enforcement

– exchange

– negotiation tools

• Integrate it into users enviroment

• Tools and measurements to determine effectiveness


Integrated Research Agenda• Advance the State of Computational Science and

Engineering– immerse a community in a computationally rich environment with

the right tools: algorithms, programming & system support

– Path to exploiting novel techniques and technology

• Explore design techniques for robust large-scale distributed systems– economic (or ecologic) approach

• Explore new ways of interacting with information– large paste-ups, two hands, sketching, 3D collaboration

• Investigate new technology– SMP nodes, gigabit Ethernet, SANs, VIA

– NT, dCOM, Java beans, directory services

– workbench displays, 3D icons, haptics, position sensors


Perspective

• Highly leveraged investment in a large scale infrastructure for studying problems of scale

• Deep commitment across the campus

• Sense of ownership and participation

• Rich research agendaOverall 11M$ Budget

Breakdownin K$

6,000

1,850

1525

1000

400

300

600

Intel

NSF

UCB

IBM

Microsoft

Sun

Synoptics

Documents

SimMillennium Project Overview David E. Culler Computer Science Division U.C. Berkeley NSF Site Visit March 2, 1998