Grid Resource Management: Challenges, Approaches, & Solutions Dr. Rajkumar Buyya Cloud Computing and Distributed Systems (CLOUDS) Lab. The University of.

Grid Resource Management: Challenges, Approaches, &

Solutions

Dr. Rajkumar BuyyaCloud Computing and Distributed Systems (CLOUDS)

Lab. The University of MelbourneMelbourne, Australiawww.cloudbus.org

2

Agenda

Grid Challenges Revisited Foundations of Resource Management

Challenges Decentralised Grid Scheduling Approach

Service Oriented-Grid Architecture Market-Oriented Grid Middleware On-Demand Assembly of UtilityGrids Summary

Scheduling Economics

Grid

Grid Economy

3

Grid Challenges: RM and Scheduling

Security

Resource Allocation & Scheduling

Data locality

Network Management

System Management

Resource Discovery

Uniform Access

Computational Economy

Application Construction

4

Open-Source Grid Middleware Projects

5

Driving Theme:Community vs. Utility Grids

Type

Feature

Community Grids Utility Grids

User QoS Best effort Contract/SLA

Service Pricing

Not considered /

free access

Usage, QoS level, Market supply and demand

Example Middleware

Globus, Condor, OMII, Unicore

Nimrod-G, Gridbus, & many inspired efforts (CatNets, Sun Grid Market, IBM..)

6

Why “integrate” Scalable Architecture, Business Models, and

Optimal Allocation

WWG

Pushes Grid computing into mainstream

computing

Gridbus

Foundations of Grid Resource Management

8

Resource Management Systems (RMSs): General Goals

Manage the Supply and Demand for Resources Allocate Resources such that:

They are allocated on fairly They are effectively utilised Most users are satisfied High priority jobs are given prominence

In a wide-area systems such as Grids: Additionally, we need to make sure that Resource

Providers are given appropriate “incentive” for their contribution and to ensure sustained resource sharing.

Therefore, Resource Management is a Challenging Task due to

their complex characteristics and goals.

9

Resources in Grid Environment: Characteristics

Autonomous Each have their own resource allocation policy no central control

Heterogeneous and substrate: Each resource can be different – SMPs, Clusters,

Linux, UNIX, Windows, Intel, etc. Resource owners have their own policies or

scheduling mechanisms. Varying Availability

Resource allocation/co-allocation challenge The amount of resource available various with time

10

Resources in Grid Environment: Characteristics

Size (large number of nodes, providers, consumers) Heterogeneity of:

resources (PCs, Workstatations, clusters, and supercomputers)

fabric management systems (single system image OS, queuing systems, etc.)

fabric management polices applications (scientific, engineering, and commerce) application requirements (CPU, I/O, memory, and/or

network intensive) demand patterns (peak, off-peak)

Geographic distribution and different time zones Differing goals (producers and consumers have different

objectives and strategies) Unsecure and Unreliable environment

11

RMS Architecture Alternatives

R1

R2

Rn

. ..

...

R

R1

Rn

..

R

R

R1

Rn

..

…….

Centralized (Single Resource)

Decentralized (Hierarchical)

(Multiple Domains)

Centralized (Multiple Resources)

(Single/Multiple Domains)

Decentralized(Self coordinated or Job Pool)

(Multiple Domains)

R

Scheduler Model Scheduler Architecture Example System

Unix, Linux,Windows OS

Multi-clusters/Grid Systems:

AppLes, SGE-E Nimrod-G,

Gridbus BrokerEnterprise Grids:

Alchemi, SETI@Home

1. Cooperative Clusters (exchange

workload)2. P2P systems (no exchange of jobs)

Cluster systems:PBS, LSF,

SGE, Condor

(Resource Broker orGrid Scheduler)

(Job Pool)

(Jobs) (Queue)

(Local Scheduler)

12

Centralized Vs. Decentralized Resource Management in Grids?

Traditional systems use centralised policy that need complete state-information and common fabric management policy or decentralised consensus-

based policy. Due to too many heterogenous parameters in the Grid it is

impossible to define/get: system-wide performance matrix and common fabric management policy that is acceptable to all.

Therefore, “decentralised” approach towards management of resource is advocated.

Question is: Market-oriented approaches found to be very effective in

managing complexities of decentralisation present in “human” economy, can they be applied to Grid Resource Management?

A Case for Economy-based Grid Resource Management

Service-Oriented Grid Architecture

14

What do Grid players want?

Grid Consumers Execute jobs for solving varying problem size and

complexity Benefit by utilizing distributed resources wisely Tradeoff timeframe and cost

Strategy: minimise expenses

Grid Providers Contribute resources for executing consumer jobs Benefit by maximizing resource utilisation Tradeoff local requirements & market opportunity

Strategy: maximise return on investment

15

What do Grid players want & require?

Grid Service Consumers (GSCs): - minimize expenses, meet QoS How do I express QoS requirements ? How do I trade between timeframe & cost ? How do I discover services and map jobs to meet my QoS needs? How do I manage Grid dynamics and get my work done? …

Grid Service Providers (GSPs):– maximise ROI How do I decide service pricing models ? How do I specify them ? How do I translate them into resource allocations ? How do I enforce them ? How do I advertise & attract consumers ? How do I do accounting and handle payments? …

They need mechanisms, tools and technologies that help them in value expression, value translation, and value enforcement.

16

Principle 1: Service Oriented Architecture (SOA)

A SOA is a contractual architecture for offering and consuming software as services.

There are four entities that make up an SOA service provider, service registry, and service consumer (also known as service requestor).

The functions or tasks that the service provider offers, along with other functional and technical information required for consumption, are defined in

the service definition or contract.

provider

registry

consumer

contract

17

Principle 2: Market-Oriented (Grid) Computing- (a) Sustained Resourced Sharing and (b)

Effective Management of Shared Resources

Grid Economy

18

Market-based Systems = Self-managed and Self-regulated systems.

Manage Complexity Supply and

Demand Enhance Utility

1

32

penalty

19

Cost Matrix @ Grid site X

Non-uniform costing Encourages the use

of local resources first

Real accounting system can control machine usage

11 33

22 11User 5User 5

Mach

ine 1

Mach

ine 1

User 1User 1

Mach

ine 5

Mach

ine 5

Resource Cost = Function (cpu, memory, disk, network, software, QoS, current demand, etc.)

Simple: price based on peaktime, offpeak, discount when less demand, ..

20

Market-based Computing Systems Need to Support:

To enable users (GSPs and GSCs) to realise economic value, market-based systems need to provide mechanisms for:

Value Expression a means to express their requirements, valuations, and

objectives Value Translation

scheduling policies to translate them to resource allocations

Value Enforcement mechanisms to enforce the selection and allocation of

differential services, and dynamic adaptation to changes in their availability at runtime

Market mechanisms, accounting and payment, Reservation of resources.

21

Grid Node N

Service-Oriented Grid Architecture

Grid Service Consumer

Pro

gra

mm

ing

En

viro

nm

ents

Grid Resource Broker

Grid Service Providers

Grid Explorer

Schedule Advisor

Trade Manager

Job ControlAgent

Deployment Agent

Trade Server

Resource Allocation

ResourceReservation

R1

Misc. services

Information Service

R2 Rm…

Pricing Algorithms

Accounting

Grid Node1

…

Core Middleware Services

…

…

HealthMonitor

Grid Market Services

JobExec

Info ?

Secure

Trading

QoS

Storage

Sign-on

Grid Bank

Ap

pli

cati

on

s

Data Catalogue

22

Market-Oriented Grid Software: A union of Gridbus and other

technologies

AIXSolarisWindows Linux

.NETGridFabricSoftware

GridApplications

Core GridMiddleware

User-LevelMiddleware

GridBank

Grid Exchange & Federation

JVM

Grid Scheduling:

Task, Parametric, and Components Programming

Gridbus Resource Broker

MPI

Condor SGE TomcatPBS

Aneka

Workflow APIs

IRIX OSF1 Mac

Libra

Globus Unicore ……Grid

MarketDirectory

PDB

CDB

Worldwide Grid

GridFabricHardware

Grid PortalsScience Commerce Engineering ……Collaboratories

……

Grid Storage Economy

Gri

d E

con

om

y

NorduGrid XGrid

ExcellGrid

Grid Workflow Engine

APIs/Tools:

23

On Demand Assembly of Services in Market-Oriented Grid Environments

ASP Catalogue

Grid Info Service

Grid Market Directory

GSP(Accounting Service)

GridbusGridBank

GSP(e.g., UofM)

PEGSP

(e.g., VPAC)

PE

GSP(e.g., IBM)

CPUorPE

Grid Service (GS)

(Globus)

Aneka

GS

GTS

Cluster Scheduler

Job

8

GridResource Broker

2

Visual Application Composer

Application CodeExplore

data1

36

45

Resu

lts9 7

Results+

Cost Info

10

11

Bill

12Data Catalogue

24

On Demand Assembly of Services in Market-Oriented Grid Environments:

Putting Them All Together

ASP Catalogue

Grid Info Service

Grid Market Directory

GSP(Accounting Service)

GridbusGridBank

GSP(e.g., UofM)

PEGSP

(e.g., VPAC)

PE

GSP(e.g., IBM)

CPUorPE

Grid Service (GS)

(Globus)

Aneka

GS

GTS

Cluster Scheduler

Job

8

GridResource Broker

2

Visual Application Composer

Application CodeExplore

data1

36

45

Resu

lts9 7

Results+

Cost Info

10

11

Bill

12Data Catalogue

25

Summary

Resource management is a complex undertaking as systems need to be adaptive, scalable, competitive,…, and driven by QoS.

A “computational economy”-based resource allocation helps in regulating the supply-and-demand for resources.

The use of economic paradigm for resource management and scheduling is essential for pushing Grids into mainstream computing.

Next: Nimrod-G Broker, Globus middleware, Aneka Cloud Middleware