Resource Brokering in the PROGRESS Project

Juliusz Pukacki

Grid Resource Management Workshop, 21-22 October 2003

Progress Project

Duration: December 2001 – May 2003Deployment: June 2003-December 2003Project Partners

SUN Microsystems PolandPSNC IBCh PoznańCyfronet AMM, KrakówTechnical University Łódź

Co-funded by The State Committee for Scientific Research (KBN) and SUN Microsystems Poland

PROGRESS Goals

Provide computational portal for scientistsProvide easy to use user interface

Provide Grid middleware infrastructure

Make use of distributed cluster of SUN SF6800 is a pilot testbed installation

Make use of PIONIER network infrastructure

PROGRESS Testbed

Cluster of 80 processors

Networked Storage of 1,3 TB

Software: ORACLE, HPC Cluster Tools, Sun ONE, Sun Grid Engine

Architecture - overview

PORTAL

HPCRESOURCES

GRIDMANAGEMENT

SYSTEM

GRID SERVICEPROVIDER

4-tier new4-tier newgrid-portal environmentgrid-portal environment

PORTAL

HPCRESOURCES

GRIDMANAGEMENT

SYSTEM

3-tier classical3-tier classicalgrid-portal grid-portal

environmentenvironment

Grid Broker - goals

Provide universal mechanism of resource management for Grid environment

Make use of common lower level Grid infrastructure (Globus)

Service oriented approach

Well-defined universal interface, implemented in common technology (Web Services)

Modular architecture

Grid Broker - functionality

Ability to choose the best resource for the job execution, according to Job Description and chosen mapping algorithm;Ability to submit the GRMS Task according to provided Job Description;Ability to submit a set of tasks with precedence constraints (sequence/parallel structure)Provides information about the Task status;Provides other information about Tasks (name of host where the Task is/was running, start time, finish time);

Grid Broker – functionality (2)

Provides list of resources that fullfills requirements provided in Job Description);Provides a list of Tasks submitted by given user;Ability to transfer input and output files (GridFTP, GASS, Data Management System);Ability to cancel the Task;Provides notification mechanism

Grid Broker - overview

Broker

Progress Grid Middleware

Data Management

System

•MDS•GRAM•GridFTP•GASS

Grid Infrastructure

PROGRESSPortal

User Access Layer

InformationService

Grid Broker - details

ResourceDiscovery

Broker

TaskManager

WebService

Interface

PROGRESS Portal

System LayerServices

PROGRESSServices

WorkflowMngmt.

TaskRegistry

DB

Grid Broker - modules

Web Service InterfaceProvides GSI enabled web service interface for Clients (PROGRESS Portal)Supports callback communication to Client (notification service)

Workflow Management ModuleCreates workflow graph of tasks from Job DescriptionPuts tasks to databaseControls of tasks execution according to precedence constraints

Grid Broker – modules (2)

Broker ModuleManages process of task submittion

Chooses the best resources for task execution

Transfers executable to chosen resource

Transfers input and output files

Resource Discovery ModuleFinds resources that fullfills requirements described in Job Description

Provides information about resources, required for task scheduling

Grid Broker – modules (3)

Task Manager ModuleAbility to check current status of task

Ability to cancel running task

Monitors for status changes of running task

Task RegistryStores information about Task execution (start time, finish time, machine where executed, current status, Job Description)

Job Description - XRSL

Provides ability to express precedence constraints of Task

Grid Broker’s Job contains one or more Tasks

Tasks are grouped into sequence or parallel execution

Job Description - schema

Parallel/sequence structure

Job1

single

Task1

Job3

parallel

Job2

sequence

Task2 Task3Task1 Task1

Task2

Task3

Task2a Task2b Task3c

Task2a

Task2b

Task2c

Job Description - Task

Task executable file location arguments file argument (files which have to be present in working directory of running executable) environment variables standard input standard output standard error

Job Description - Task

Resource requirements name of host for job execution operating system required local resource management system minimum memory required minimum number of cpus required minimum speed of cpu other parameter passed directly to Globus GRAM

XRSL - example<?xml version="1.0" encoding="UTF-8"?><job id="Application">

<user dn="dc=pl, dc=poznan, dc=man, uid=pukacki"/><task id="Task1">

<executable name="computation" dmsid="12345" type="single" count="1">

<arguments><value>4</value><file id="772" type="in">mydata.dat</file>

</arguments><environment>

<stdin dmsid ="772"></stdin><stdout dmsid="774"/><stderr dmsid="775"/><variable name="CONST">5</variable>

</environment></executable><resources>

<resource type="memory" value="20"/></resources>

</task></job>

XRSL – example (2)<?xml version="1.0" encoding="UTF-8"?><job id="Application">

<user dn="dc=pl, dc=poznan, dc=man, uid=pukacki"/><sequence id="seq1">

<task id="Task1"><executable name="computation" dmsid="12345">

<arguments><value>4</value>

</arguments><environment>

<stdout dmsid="774"/></environment>

</executable></task><task id="Task2">

<executable name="computation" dmsid="8888"><arguments>

<value>4</value></arguments><environment>

<stdout dmsid="774"/></environment>

</executable></task>

</sequence></job>

Thank you