MATLAB DISTRIBUTED COMPUTING SERVER 4 - SYSTEM ADMINISTRATORS GUIDE Instrukcja Użytkownika

MATLAB®Distributed Computing Engine 3System Administrator’s Guide

1 IntroductionWhat Are the Distributed Computing Products?In this section...“Overview” on page 1-2“Determining Product Installation and Versions” on p

What Are the Distributed Computing Products?MATLAB WorkerSchedulerorJob ManagerMATLAB ClientDistributedComputingToolboxMATLAB DistributedComputing Eng

1 IntroductionToolbox and Engine ComponentsIn this section...“Job Managers, Workers, and Clients” on page 1-4“Third-Party Schedulers” on page 1-6“Comp

Toolbox and Engine ComponentsWorkerSchedulerorJob ManagerClientWorkerWorkerClientJobAll ResultsJobAll ResultsTaskResultsTaskResultsTaskResultsInteract

1 IntroductionThird-Party SchedulersAs an alternativetousingtheMathWorksjobmanager,youcanuseathird-party scheduler. This could be Windows CCS, Platfor

Toolbox and Engine ComponentsIf you have a large cluster, you probably already have a sche duler. Consultyour MathWorks repres entative if you have qu

1 IntroductionUsing Distributed Computing ToolboxA typical Distributed Computing Toolbox client session includes the followingsteps:1 Find a Job Manag

2Network AdministrationThis chapter provides information useful for network administration ofDistributed Computing Toolbox and MATLAB Distributed Comp

2 Network AdministrationPreparing for Distributed ComputingIn this section...“Before You Start” on page 2-2“Planning Your Network Layout” on page 2-2“

Preparing for Distributed ComputingSession Product ProcessesClient Distributed ComputingToolboxMATLAB with toolboxWorker MATLAB DistributedComputing E

How to Contact The MathWorkswww.mathworks.comWebcomp.soft-sys.matlab Newsgroupwww.mathworks.com/contact_TS.html Technical [email protected]

2 Network AdministrationSecurity ConsiderationsThe distributedcomputing products do not provide any security measures.Therefore, you should be aware o

Installing and ConfiguringInstalling and ConfiguringTo find the most u p-to-date instructions for installing and configuring thecurrent or past versio

2 Network AdministrationShutting Down a Job Manager ConfigurationIn this section...“UNIX and Macintosh” on page 2-6“Windows” on page 2-8If you are don

Shutting Down a Job Manager ConfigurationIf you have more than one worker session running, you can stop each ofthem individually by host and name.stop

2 Network AdministrationWindowsStopping the Job Manager and Workers1 To shut down the job manager, enter the commandscd matlabroot\toolbox\distcomp\ b

Shutting Down a Job Manager Configurationcd matlabroot\toolbox\distcomp\ binmdce stopIf you plan to uninstall MATLAB Distributed Computing Engine from

2 Network AdministrationCustomizing Engine ServicesIn this section...“Defining the Script Defaults” on page 2-10“Overriding the Script Defaults” on pa

Customizing Engine ServicesSetting the UserBy default, the job manager and worker services run as the user who startsthem. You can run the services as

2 Network Administration• matlabroot\toolbox\distcomp\bin\mdce_def.bat (Windows)•matlabroot/toolbox/distcomp/bin/mdce_def.sh (UNIX or Macintosh)Before

Accessing Service Record FilesAccessing Service Record FilesIn this section...“Locating Log Files” on page 2-13“Locating Checkpoint Directories” o n p

Revision HistoryNovember 2005 Online only New for Version 2.0 (Release 14SP3+)December 2005 Online only Revised for Version 2.0 (Release 14SP3+)March

2 Network AdministrationLocating Checkpoint DirectoriesCheckpoint directories contain information related to persistence data, whichthe engine service

Accessing Service Record FilesPlatform File LocationWindows On Windows systems, the defaultlocation of the checkpoint directories is<TEMP>\MDCE\

2 Network AdministrationTroubleshootingIn this section...“License Errors” on page 2-16“Verifying Multicast Communications” on page 2-18“Memory Errors

Troubleshooting• If you receive this error when starting a worker with the DistributedComputing Engine- You may be calling the startworker command fro

2 Network AdministrationHave your MATL AB administrator ve rify that the license manageris running and validate network services.For more infor mation

TroubleshootingInside MATLAB, the class would be used as follow s.m = com.mathworks.toolbox.distcomp.test.Multicast Tester('239.1.1.1', 9999

2 Network AdministrationRequired PortsUsing a Job ManagerBASE_PORT. The ports required by the job manager and all workers arespecified and described i

Troubleshooting3 On the Edit menu, click New, and then add the following registry entry:Value Name: MaxUserPortValue Type: DWORDValue data: 65534Valid

2 Network Administration2-22

3Control Scripts — ByCategoryMDCE Control (p. 3-2)Control mdce serviceJob Manager Control (p. 3-2)Control job managerWorker Control (p. 3-2)Control MA

3 Control Scripts — By CategoryMDCE Controlmdce Install, start, stop, or uninstall mdceservicenodestatus Status of MDCE processes runningon nodeJob Ma

4Control Scripts —Alphabetical List

mdcePurpose Install, start, stop, or uninstall mdce serviceSyntax mdce installmdce uninsta llmdce startmdce stopmdce consolemdce restartmdce ... -md c

mdcemdce ... -mdcedef <mdce_defaults_file> uses the specifiedalternativemdcedefaultsfileinsteadoftheonefoundinmatlabroot/toolbox/distcomp/bin.md

nodestatusPurpose Status of M DCE processes running on nodeSyntax nodestatusnodestatus - flag sDescription nodestatus displays the status of the mdce

nodestatusExamples Display basic information about the mdce processes on the local host.nodestatusDisplay detailed information about the status of the

star tjobmanagerPurpose Start job manager processSyntax startjobmanagerstartjobmanager -flagsDescription startjobmanager starts a job manager process

star tjobmanagerFlagOperation-multicastOverrides the use of unicast tocontact the job manager lookupprocess. It is recommendedthat you not use-multica

star tjobmanagerStart the job manager MyJ obManager on the host JMHost.startjobmanager -name MyJobManage r -remotehost J MHostSee Also mdce, nodestatu

startworkerPurpose Start MATLAB worker sessionSyntax startworkerstartworker -flagsDescription startworker starts a MATLAB w orker p rocess under the m

ContentsIntroduction1What Are the Distributed Computing Products? ... 1-2Overview... 1-2Determining Product In

star tworkerFlagOperation-jobmanagerhost <jmhostname>Specifies the host on which thejob manager is running by using-jobmanagerhost.Theworkerwill

startworkerFlagOperation-baseport <port _nu mber>Specifies the base port that themdce service on the remote hostis using. You only need to speci

star tworkerSee Also mdce, nodestatus, startjobmanager , stopjobmana ger, stopwor ker4-12

stopjobmanagerPurpose Stop job manager processSyntax stopjobmanagerstopjobmanager -flagsDescription stopjobmanager stops a job manager that is running

stopjobmanagerExamples Stop the job manager MyJobManager on the local host.stopjobmanager -name MyJobManagerStop the job manager MyJobManager on the h

stopworkerPurpose Stop MATLAB worker sessionSyntax stopworkerstopworker - flag sDescription stopworker stops a MATLAB worker process that is running u

stopworkerExamples Stop the worker with the default name on the local host.stopworkerStop the worker with the default name, running on the computerWor

GlossaryGlossaryCHECKPOINTBASEThenameoftheparameterinthemdce_def file that defines the locationof the job manager and worker checkpoint directories.ch

Glossarydistributed computingComputing with distributed applications, running the application onseveral nodes simultaneously.distributed computing dem

GlossaryjobThe complete large-scale operation to p erform in MATLAB, composedof a set of tasks.job managerThe MathWorks process that q ueues jobs and

Defining the Script Defaults ... 2-10Overriding the Script Defaults... 2-11Accessing Servic e Record Files..

GlossarymdceThe service that has to run on all machines before they can run a jobmanager or worker. This is the engine foundation p rocess, making sur

GlossaryschedulerThe process, either third-party or the MathWorks job manager, thatqueues jobs and assigns tasks to workers.taskOne segment of a job t

GlossaryGlossary-6

IndexIndexAadministrationnetwork 2-1Ccheckpoint directorydefinition Glo ssary-1locating 2-14CHECKPOINTBASEdefinition Glo ssary-1clean statestarting se

Indexdatabasedefinition Glo ssary-3definition Glo ssary-3logs 2-13lookup processdefinition Glo ssary-3multiple on one machine 2-10process 1-4stoppingo

Indexlicense errors 2-16memory erro rs 2-19verifying multicast 2-18Windows network installation 2-19Uusersetting 2-11Wworkerdefinition Glo ssary-5proc

Control Scripts — Alphabetical List4GlossaryIndexvii

viii Contents

1IntroductionThis chapter provides an introduction to the co nce pts and terms of DistributedComputing Toolbox and MATLAB®Distributed Computing Engine