Presentation is loading. Please wait.

Presentation is loading. Please wait.

GT4 GRAM: A Functionality and Performance Study Stuart Martin, Martin Feller Computational Institute, University of Chicago & Argonne National Lab TeraGrid.

Similar presentations

Presentation on theme: "GT4 GRAM: A Functionality and Performance Study Stuart Martin, Martin Feller Computational Institute, University of Chicago & Argonne National Lab TeraGrid."— Presentation transcript:

1 GT4 GRAM: A Functionality and Performance Study Stuart Martin, Martin Feller Computational Institute, University of Chicago & Argonne National Lab TeraGrid 2007 Madison, WI

2 2 Contributors / Collaborators l UC/ANL –Ian Foster –Peter Lane –Jarek Gawor –Ravi Madurri –Rachana Ananthakrishnan

3 3 GRAM - Basic Job Submission and Control Service l A uniform service interface for remote job submission and control –Includes file staging and I/O management –Includes reliability features –Supports basic Grid security mechanisms –Asynchronous monitoring –Interfaces with local resource managers, simplifies the job of metaschedulers/brokers l GRAM is not a scheduler. –No scheduling –No metascheduling/brokering

4 4 GRAM Versions in GT4 l GRAM2 (Pre-WS GRAM) –Proprietary Protocol based implementation –Gatekeeper and Job Manager l GRAM4 (WS GRAM) –Web Services-based implementation –Managed Job Factory Service (MJFS) –Managed Executable Job Service (MEJS)

5 5 Comparison l Functionality –Security –File Staging –General l Performance –Concurrent jobs –Sequential jobs

6 6 Security Functional Comparisons

7 7 Privilege Limiting Model l GRAM must be able to start jobs submitted by remote users under different user ids. It must execute some code as root –GRAM2: Entire gatekeeper runs as root –GRAM4: Service with sudo privs >non-root container account requires sudo to invoke operations as other users

8 8 Authentication l A client can authenticate with GRAM with a variety of protocols –GRAM2: TLS (only) –GRAM4: TLS, Message Level Security >Message-level WS-Security >Channel-level WS-SecureConversation >Choice for which to support in each deployment

9 9 Credential Delegation l Needed by GRAM or the users applications to do file staging or other grid operations –GRAM2: Yes, Required >Clients must delegate from client to service on every request –GRAM4: Yes, Optional >Clients can choose and delegate when necessary

10 10 Credential Refresh l Credentials have a lifetime and may expire before a job has completed execution –GRAM2: Yes –GRAM4: Yes >A client can query for information about the WS Resource of the delegated credential >Remaining lifetime

11 11 Share credential delegation among jobs l When repeatedly interacting with the same GRAM service, a client may want to delegate once and share the delegation among multiple jobs –GRAM2: No –GRAM4: Yes >Refreshing a credential in the delegation service that was shared among multiple job submission will results in a refresh for each job

12 12 Authorization Callouts l Following authentication, GRAM checks to see if the request should be authorized. For example, a gridmap file acting as an access control list –GRAM2: Yes - single PDP callout –GRAM4: Yes - Multiple PDP callout chain >Allows for richer policies l Parse VOMS attributes l Use attributes in policy evaluations l Site level black lists

13 13 File Management Functional Comparisons

14 14 File Staging l Job staging before and after the users job is executed –GRAM2: Yes –GRAM4: Yes

15 15 File staging retry policy l If a file staging operation fails, it may be non-fatal and retry may be desired –GRAM2: None –GRAM4: RFT Supported >Server defaults for all transfers can be configured >Defaults can be overridden for a specific transfer

16 16 Incremental output staging streaming l It can be useful to obtain access to data produced by a program as it executes. –GRAM2: stdout/stderr only –GRAM4: stdout/stderr and any file >A client can stream files via the service-side GridFTP server. This is what globusrun-ws does for stdout and stderr streaming.

17 17 Standard input access l The contents of a file can be passed to the jobs standard input –GRAM2: Yes –GRAM4: Yes

18 18 Throttle staging work l A GRAM submission that specifies file staging imposes load on the service node executing the GRAM service. –GRAM2: No –GRAM4: Yes >GRAM is configured for a maximum number of worker threads and thus a maximum number of concurrent staging operations.

19 19 Load balance staging work l Allow staging work to be load balanced among a set of service hosts –GRAM2: No –GRAM4: Yes >Staging work can be distributed over several service nodes. For example, a separate GridFTP server can be configured for each LRM type or file system paths.

20 20 General Functional Comparisons

21 21 Access protocol l Protocol used to interact with the service –GRAM2: proprietary HTTP –GRAM4: Web Service SOAP >Standards based l WSDL l Client tooling

22 22 Job Description Language l The mechanism for specifying job directives. –GRAM2: RSL >Custom string-based language –GRAM4: JDD >Job description document (JDD) XML-based version >Initial prototype of OGFs JSDL specification

23 23 Extensible Job Description Language l A mechanism for passing extensions through GRAM to underlying local resource managers –GRAM2: Yes –GRAM4: Yes

24 24 Local Resource Manager Interface l The GRAM interface to the LRM to submit, monitor, and cancel jobs. –GRAM2: Perl scripts –GRAM4: Perl scripts + SEG >Scheduler Event Generator (SEG) provides efficient monitoring between the GRAM service and the LRM for all jobs for all users

25 25 Local Resource Managers l Supports a range of LRMs - PBS, LSF, Condor, Fork, … –GRAM2: Yes –GRAM4: Yes

26 26 Fault Tolerance l GRAM can recover from a container or host crash. Upon restart, GRAM will resume processing of the users job submission –GRAM2: Yes - Client initiated >Processing resumes for a single job after the client has restarted the job manager service process –GRAM4: Yes - Service initiated >Processing resumes for all jobs once the service container has been restarted

27 27 State Access: Push (subscription) l Allow clients to request notifications for state changes –GRAM2: Yes - callbacks –GRAM4: Yes - WS Notifications >Clients can subscribe for notifications to the job status resource property

28 28 State Access: Pull l Allow clients to get the state for a previously submitted job –GRAM2: Yes >The service defines a proprietary operation to get the job state. –GRAM4: Yes >The service defines a WSRF resource property that contains the value of the job state. A client can then use the standard WSRF getResourceProperty operation.

29 29 Audit Logging l Allow an audit records to be inserted into an audit DB when a job completes –GRAM2: Yes –GRAM4: Yes >An enhancement was contributed by Gerson Galang (APAC) to insert the record at the beginning of the job and to update the audit record after submission and again at job end.

30 30 At Most Once Job Submission l A simple request-reply job submission protocol has the problem that if the reply message is lost, a client cannot know whether a job has been started. Measures need to be taken to ensure that the same job is not submitted twice. –GRAM2: Yes - 2-phase commit >Requires an extra round trip, plus a delay on the service to begin processing –GRAM4: Yes - UUID on create >The client supplies a client-created unique ID (UUID) and the GRAM4 service guarantees not to start a job with a duplicate ID

31 31 Job Cancellation l Allow a job to be cancelled –GRAM2: Yes >Proprietary operation –GRAM4: Yes >WSRF standard Destroy operation

32 32 Job Lifetime Management l Allow a client to control when a jobs state is cleaned up –GRAM2: Yes >Implements a set of job directives and operations –GRAM4: Yes >Standard WS-ResourceLifetime operations

33 33 Maximum Active Jobs l The Maximum number of jobs that the service can manage –GRAM2: ~250 >Due to each job Job Manager process querying the LRM separately –GRAM4: 32,000 >Limited by the number of directories that can be created in a directory

34 34 Parallel Job Support l Support for MPI jobs jobtype = MPI –GRAM2: Yes –GRAM4: Yes

35 35 MPICH-G Support l Support for multi-site MPI –GRAM2: Yes >Client-side DUROC and service-side DUCT service –GRAM4: Yes >Multi-job and rendezvous Web Services >MPIg support coming soon

36 36 Basic Execution Service (BES) Interface l Support for OGSA BES for job submission –GRAM2: No –GRAM4: Prototyped >Working on plans to initially support JSDL with the current GRAM4 port type, then add support for BES too

37 37 Performance Comparisons

38 38 Concurrent Jobs (as in paper) Stage In Stage Out File Clean Up Unique Job Dir GRAM2GRAM4 None No 25522100 1X10KB No 26083779 1X10KB Yes 26985695 Average seconds per 1000 jobs Condor-g to GRAM to Condor LRM

39 39 Concurrent Jobs (as will be in GT 4.0.5) Stage In Stage Out File Clean Up Unique Job Dir GRAM2GRAM4 None No 25522176 1X10KB No 26082147 1X10KB Yes 26982254 Average seconds per 1000 jobs Condor-g to GRAM to Condor LRM

40 40 Improving performance for staging jobs l Adding local method call mechanism for general use in Java WS Core (4.0.5) –GRAM is doing this with RFT –Any service which calls another in-process service could make similar modifications for local calls and likely benefit from improved performance l Adding caching of the GridFTP server connections in RFT (4.0.6)

41 41 Sequential Jobs Delegation Stage In Stage Out GRAM2GRAM4 None N/A1.70 Per JobNone 1.073.53 Per Job1X10KBNone1.785.57 Shared1X10KBNoneN/A5.41 Per Job1X10KB 2.449.08 Shared1X10KB N/A7.91 Average seconds per job (Fork)

42 42 Sequential Jobs Delegation Stage In Stage Out GRAM2GRAM4 None N/A1.46 Per JobNone 1.073.42 Per Job1X10KBNone1.783.46 Shared1X10KBNoneN/A3.51 Per Job1X10KB 2.445.25 Shared1X10KB N/A3.67 Average seconds per job (Fork)

43 43 For More Information l Stuart Martin - l Martin Feller -

Download ppt "GT4 GRAM: A Functionality and Performance Study Stuart Martin, Martin Feller Computational Institute, University of Chicago & Argonne National Lab TeraGrid."

Similar presentations

Ads by Google