Transcript No Slide Title

http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Towards a High Performance
Extensible Grid Architecture
Klaus Krauter
Muthucumaru Maheswaran
{krauter, mahes}@cs.umanitoba.ca
Computer Science Department
University of Manitoba
Winnipeg Manitoba
http://www.cs.umanitoba.ca/~anrl
Outline
Advanced
Networking
Research
Lab
• Grid Computing Issues



Network computing environment
Scalability, Extensibility, and Adaptability
Quality of Service
• Grid Models


Resource Management Techniques
Application Execution Models
• Grid Architecture
• Example Applications

Compiling, Numerical Processing, Grid Aware
Application
• Related Work
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Computing Issues
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Network Computing Environment
• Heterogeneous Nodes


Autonomous administration domains with
different resource management policies
Servers, network devices, workstations,
PDA, etc.
• Connected by Communication Links

Support differentiated service levels
• Use native operating system services


Does not replace existing scheduling and
resource control mechanisms
Native operating system is a Grid device
driver
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Scalability
• Target Size


Hundreds to Millions of nodes
Different platforms for different scale Grids
• Global resource management protocols


Fixed format messages
Ability to locally tune protocol performance
parameters to match local infrastructure
and administrative policy
• Local policies for resource management

Scheduling, Quality of Service, Tolerance to
faults
http://www.cs.umanitoba.ca/~anrl
Extensibility and Adaptability
Advanced
Networking
Research
Lab
• Extensible resource protocol content

Fixed message framework with structured
extensibility (XML like)
• Extensible resource management protocol
processing


Message content extensions are processed by
extension modules
Modules are dynamically loaded and register content
identifiers
• Variability

Multiple different implementations of the resource
protocols
• Adaptability


Nodes and resources enter and leave the grid
continuously
Fault tolerance by resource replication
http://www.cs.umanitoba.ca/~anrl
Quality of Service
Advanced
Networking
Research
Lab
• Not restricted to end-to-end network

Processor, memory, I/O also need to support QoS
specifications
• Co-allocation and Co-reservation

Allocation and scheduling need to take into account
QoS given to other jobs already in the Grid
• Providing Service Level Agreements


Aggregate performance levels or on a per job basis?
Site autonomy and resource control restricts the
ability to provide guarantees
• Applications should be able to negotiate QoS with
the Grid
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Models
http://www.cs.umanitoba.ca/~anrl
Resource Management
Techniques
Advanced
Networking
Research
Lab
• Super Scheduler


Hierarchy of cooperating schedulers
Issues: Co-allocation
• Market Based


Auctioning for resources
Issues: Price management and co-allocation
• Resource Discovery



Resource attribute and status in a distributed
database
Centralized, Agent based, or Hybrid
Issues: devise highly distributed, scalable, fault
tolerant schemes
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Application Execution Models
• Legacy application




Native OS resource and scheduling, implicit QoS
Use external resource description language
Modify native OS and service libraries and infer
resource requirements and QoS
Recompile with Grid aware compiler that inserts
specialized Grid code
• Grid Aware application


Use specialized Grid API
First “applications” will be compilers, service libraries
(MPI, PVM), Grid workbenches and monitoring tools
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Aware Applications
Libraries/Components
Grid Workbench
Application Source
Execution Plan
Grid Aware Compiler
Application Information
Service
QoS Broker
Resource Specific
Optimizer
Performance
Monitor
Grid
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Non-Grid Aware Applications
Grid Workbench
Execution Plan
Application
Information
Service
QoS Broker
Resource Specific
Optimizer
Performance
Monitor
Grid
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Architecture
http://www.cs.umanitoba.ca/~anrl
Design Approach
Advanced
Networking
Research
Lab
• Layered



Grid Kernel
Grid Core Services
Grid toolkits, workbenches, and user
interfaces
• Fully distributed peer-to-peer model


No centralized information servers
Implementations free to use specialized
servers
• Minimal configuration

Use Service Location Protocol like service
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Kernel Architectural
Principles
• Functions that use the services are aware of the
distributed environment
• No guarantees made about reliability of nodes or
links
• Operate on all types of heterogeneous nodes
using minimal resources
• Services will be implemented using native OS with
minimal changes to trusted computing base
• Provide uniform extensible API and services across
all nodes
• Provide resource management mechanisms but do
not implement resource management policies
http://www.cs.umanitoba.ca/~anrl
Grid Architecture
Advanced
Networking
Research
Lab
Legacy Applications
and Libraries
Grid Applications
Grid Application and
Problem Solving Environments
Grid User Interfaces
Programming Toolkits
Grid Kernel
Security
Management
Dynamic
Extension
Dynamic
Extension
Resource
Optimizer
Recovery
Services
Location
Services
Scheduler
QoS Broker
.
.
.
Accounting
Grid Core Services
Module to Kernel Interface Layer
Resouce
Protocol Processing
Module
Management
Namespace
Management
Grid Kernel to Operating System Interface Layer
Native Operating System and Services
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Layers and Core Services
Location Services
Node 6
Recovery
Services
QoS Broker
Node 5
Node 4
Accounting
Naming
Node 2
Node 3
IP Traffic Optimizer
Service Stubs
Resident Services
Grid Kernel
Node 1
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Grid Example Applications
http://www.cs.umanitoba.ca/~anrl
Applications
Advanced
Networking
Research
Lab
• Compiling



Ensure similar compiler and libraries are used on all
nodes
Compute how long to transfer and compile
Perform deadline scheduling
• Legacy Numerical Processing



Dynamically linking of Grid code, variable QoS for
job steps
Describe network QoS requirements or infer
dynamically
Much further research required
• Collaborative Research Workbench



Negotiate video bandwidth required
Query if a simulation can be run and completed
quickly, or schedule it later
Different GUI depending on resources nearby to a
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Related Work
http://www.cs.umanitoba.ca/~anrl
Related Work
Advanced
Networking
Research
Lab
• Application Enabling Systems


Provide tools to allow applications to access
globally distributed resources in an
integrated fashion
ATLAS, Globe, Globus/GUSTO, Legion,
ParaWeb, Symera
• User Access Systems


Provide end users of the Grid transparent
access to geographically distributed
systems in a location independent manner
CCS, MOL, NetSolve, PUNCH
http://www.cs.umanitoba.ca/~anrl
Advanced
Networking
Research
Lab
Questions ?