TWiki> APACgrid Web>GatewayMachines>VmdetailsNg2>MdsFourConfig>GlueSchemaFields (03 Dec 2007, GersonGalang)EditAttach

Do not use this page, it has been migrated to the ARCS WIKI at http://wiki.arcs.org.au/bin/view/Main/GlueSchemaFields

GLUE Schema fields

Below is a list resource information that can be queried from MDS using the GLUE schema. The table shows the values you should expect each of the GLUE fields should have and if those fields are being published on the APAC Grid.

Site


Attribute name Attribute description Example Implemented at APAC?
Name human-readable name SAPAC Y
Description short description of the site South Australian Partnership for Advanced Computing Y
UserSupportContact   mailto:helpdesk@sapac Y
SysAdminContact   mailto:systems@sapac,grid-admin@sapac.edu.au Y
SecurityContact   mailto:systems@sapac,grid-admin@sapac.edu.au Y
Location   Adelaide, South Australia, Australia Y
Latitude   -34.919517 Y
Longitude   138.603226 Y
Web URI of the site's website http://www.sapac.edu.au Y
Sponsor VOs sponsoring the site ExampleUniversity? of Adelaide, University of South Australia, Flinders University Y
OtherInfo key value pair attribute that is used to publish other info that does not fit any attribute of the site entity ClusterStatus=Down N

Computing Element

ComputingElement? is an abstraction of a queue

Attribute name Attribute description Example Implemented at APAC?
UniqueID unique ID of this CE hydra@hydra or hydra Y
Name name of the batch queue. I usually search for the Name instead of the Unique ID. hydra@hydra or hydra Y
InformationServiceURL contact URI of the service providing status and characteristics info of this CE https://ng2.sapac.edu.au:8443/
wsrf/services/DefaultIndexService 		N
LRMSType Local Resource Management System type. In short, queuing system type. There's a list on the allowed values for this field in here Torque Y
LRMSVersion   1.1.0 Y
GRAMVersion Globus GRAM protocol version 4.0.3 N
HostName name of the machine running the service hydra.sapac.edu.au Y
GatekeeperPort It's safe to assume that all APAC partner sites are running their GT container on port 8443   N
JobManager The job manager to be used to get to this CE. jobmanager-pbs N
ContactString String specifying on how to contact the service. We're still using the GT2 format because there's no suggested way on what the format should be if the service is provided by a GT4 container. ng2.sapac.edu.au/jobmanager-pbs Y
TotalCPUs not used. Use FreeJobSlots instead   N
ApplicationDir Path of the directory available for application installation UNAVAILABLE Y
DataDir path of a shared directory available for staging in data UNAVAILABLE Y
DefaultSE Unique ID of the default StorageElement? for this CE. If a user did not specify a place to put his/her data, his/her directory on this SE will be used. hydra-s.sapac.edu.au Y
Status queue status: 'Queueing' the queue can accept job submission but cannot be served by the scheduler; 'Production' the queue can accept job submissions and is served by a scheduler; 'Closed' the queue cannot accept job submission and cannot be served by a scheduler; 'Draining' the queue cannot accept job submission, but can be served by the scheduler. Note: The MIP only publishes queues that are in the Production status 'Production' Y
RunningJobs jobs in the running state 50 Y
WaitingJobs jobs in the waiting state 100 Y
TotalJobs jobs in any state 150 Y
EstimatedResponseTime Based on the accepted jobs, estimated time to last for a new job from the acceptance to the start of its execution. This field is currently not being used on the APAC Grid. There are a few algorithms around that can be used to calculate for the value for this field. If there's really a need for this field to be published, we might have to look into those algorithms any int value in seconds N
WorstResponseTime Among the accepted jobs, the worst time from the job was accepted by the service to the start of its execution. Same as EstimatedResponseTime? , this field is not being published at the moment. any int value in seconds N
FreeJobSlots number of single-processor jobs which could be started if no other jobs are submitted and no jobs finish in the interim. This field should be used to check for the number of free CPUs in the cluster 20 Y
FreeCPUs a deprecated field in the GLUE schema. Use the FreeJobSlots? instead   N
MaxWallTime The maximum amount of wall clock time allowed to each job on this CE any integer value in minutes Y
MaxCPUTime The maximum CPU time allowed to each job on this CE any integer value in minutes Y
MaxTotalJobs The maximum allowed number of jobs in the CE 9999999 (same as saying infinity) Y
MaxRunningJobs The maximum allowed number of jobs in the running state in the CE 9999999 (same as saying infinity) Y
Priority priority given to jobs in this queue 250 Y
AssignedJobSlots Number of slots for single-processor jobs in the running state.   N

VOView

VO View of the available resources of a CE. This entity represents the view of the VOs mapped to the local unix account at the site. If there are two VOs (NGAdmin and APACAdmin) mapped to the account grid-admin at SAPAC, this is the view that members of those VOs should search for in MDS.

Attribute name Attribute description Example Implemented at APAC?
LocalID local ID of the VO /NGAdmin Y
AccessControlBaseRule set of authorizations for the CE specific for this view /NGAdmin, /APACAdmin Y
RunningJobs number of running jobs on the CE owned by the VOs   Y
WaitingJobs number of waiting jobs on the CE owned by the VOs   Y
TotalJobs number of jobs in any state on the CE owned by the VOs   Y
FreeJobSlots number of single-processor jobs that can be submitted by the user if not other jobs are submitted and not jobs finish in the interim   Y
EstimatedResponseTime EstimatedResponseTime? for this VO any integer value in seconds Y
WorstResponseTime WorstResponseTme? for this VO any integer value in seconds Y
ApplicationDir   UNAVAILABLE Y
DataDir directory where this VO member can put his files in on the DefaultSE? /data/grid/grid-admin Y
DefaultSE default storage element for this VO hydra-s.sapac.edu.au Y
FreeCPUs use FreeJobSlots? for this VO instead   N

Cluster

Cluster is a set of machines which provide computing power managed by a Local Resource Management System. If you have a number of clusters at your site that is only managed by one queueing system, those clusters will be treated as just one cluster composed of a number of subclusters. You can give this cluster (the one which is composed of a number of subclusters) any name you want. What's important is the name of the queue running under this cluster and how users can get their jobs running on the queues managed by the queueing system.

Attribute name Attribute description Example Implemented at APAC?
UniqueID unique ID of the cluster hydra.sapac.edu.au Y
Name human-friendly name of the cluster. The value of this field doesn't have to be unique. My personal preference is to search for Name instead of the UniqueID. hydra.sapac.edu.au Y
TmpDir The path of a temporary directory shared across the compute nodes. We can use the SCRATCH_DIR that Steve has specified in the file handling document. Using "~/.globus/scratch" will work on both RFT and GridFTP ~/.globus/scratch Y
WNTmpDir The path of the temporary directory local to each compute node /tmp Y

SubCluster

The SubCluster entity represents a homogeneous set of hosts with the same set of host/hardware attributes. Even if all the nodes in your cluster have the same hardware attributes, you still will need to specify a SubCluster entry for it because this is the only entity in the GLUE schema which provide hardware attributes for the Cluster.

Attribute name Attribute description Example Implemented at APAC?
UniqueID   hydra.sapac.edu.au Y
Name   hydra.sapac.edu.au Y
PhysicalCPUs Total number of real CPUs 256 Y
LogicalCPUs Effective number of CPUs in the subcluster, including the effect of hyperthreading and the effects of virtualization due to the queueing system 256 Y
OperatingSystemName try running `lsb_release -sd` Fedora Core Y
OperatingSystemRelease try running `lsb_release -sr` 4 Y
OperatingSystemVersion try running `lsb_release -sc` Stentz Y
ProcessorModel   Intel(R) Xeon(TM) CPU 3.20GHz Y
ProcessorVersion     Y
ProcessorVendor   GenuineIntel? Y
ProcessorClockSpeed CPU speed in megahertz 3200 Y
ProcessorInstructionSet   fpu tsc msr pae mce cx8 apic mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm constant_tsc pni monitor ds_cpl cid cx16 xtpr Y
ProcessorOtherDescription     N
RAMSize Amount of main memory in megabytes   Y
VirtualSize RAMSize + swap size in megabytes   Y
NetworkAdapterOutboundIP Permission for outbound connectivity, even if limited true Y
NetworkAdapterInbound Permission for inbound connectivity, even if limited false Y
ArchitecturePlatformType Platform type of the host   N
ArchitectureSMPSize number of physical CPUs per compute node 2 Y
BenchmarkSI00 SpecInt2000?   N
BenchmarkSF00 SpecFloat2000?   N
App.Soft.RunTimeEnvironment Environment variable associated to an installed software package. Used at OSG but not on the APAC Grid   N

StorageElement

StorageElement is the abstraction of a storage resource. It's a group of services, protocols, and data sources.

Attribute name Attribute description Example Implemented at APAC?
UniqueID unique identifier of the SE hydra-s.sapac.edu.au Y
Name human-friendly name of the SE hydra-s.sapac.edu.au Y
Architecture String enumeration: disk, tape, multidisk, other disk Y
SizeTotal The size in GB of the storage managed by this service 100 Y
SizeFree The available size in GB of the storage that can be used to store new files for any VO/user   Y
InformationServiceURL URL of the MDS service monitoring this service   N
SEAccessProtocol.LocalID local ID uniqueue within the specific SE instance   Y
SEAccessProtocol.Type type of the access protocol. The allowed values for this field can be found in here gsiftp Y
SEAccessProtocol.Endpoint   gsiftp://ngdata.sapac.edu.au:2811 Y
SEAccessProtocol.Version   some number Y
SEAccessProtocol.Capability array of functions supported by this access protocol. file transfer Y
SEControlProtocol.LocalID local ID   N
SEControlProtocol.Type type of the control protocol. The allowed values for this field can be found in here SRM N
SEControlProtocol.Endpoint   srm:... N
SEControlProtocol.Version     N
SEControlProtocol.Capability     N

StorageArea

A portion of the storage to which a uniform set of policies applies

Attribute name Attribute description Example Implemented at APAC?
LocalID local identifier for this area. This must be unique within the storage element data-grid-test Y
Path full path of the root directory for this area /data/grid/grid-admin Y
Type String enumeration: permanent, durable, volatile, other. Most of the storage areas on the APAC Grid are volatile. This is because of the policy that files older than 5 days will be deleted from the site's shared file system. It is the user's responsibility to copy his files to a permanent storage location when the job finish volatile Y
Quota the quota assigned in KB   N
MinFileSize minimum size in bytes for any single file   N
MaxFileSize maximum size in bytes for any single file   N
MaxData maximum amount of data in bytes that may be stored by one job   N
MaxPinDuration maximum time for a file to be pinned in seconds   N
MaxNumFiles maximum number of files that may be stored by 1 job   N
UsedSpace used space in KB   N
AvailableSpace available space in KB   Y
AccessControlBaseRule set of authorizations for the StorageArea? . This tells which VOs can use this StorageArea? /NGAdmin, /APACAdmin Y

Job

Job is the information about the job currently managed by the CE

Attribute name Attribute description Example Implemented at APAC?
LocalID job id given by the queueing system hydra.1234 N
GlobalID job id given by the Globus a string N
GlobalOwner DN of the user who submitted the job   N
ExecutionTarget hostname of the machine where the job ran   N
Status queued or running   N
SchedulerSpecific scheduler specific info   N

Service

Service captures the common attributes of Grid services

Attribute name Attribute description Example Implemented at APAC?
Name human-friendly name Hydra PBS N
Type Type of the service. There's a list on the allowed values for this field in here pbs.torque.server N
Version   1.1.0 N
Endpoint network endpoint of the service hydra.sapac.edu.au N
Status status of the service OK N
StatusInfo textual explanation for the status of the service Up and running - accepting jobs N
WSDL URI of the WSDL describing the service   N
Semantics URL of the detailed description of the service   N
StartTime timestamp of the last start time of the service   N
Owner Owner of the service. Can be one or more VOs /* N

Do not use this page, it has been migrated to the ARCS WIKI at http://wiki.arcs.org.au/bin/view/Main/GlueSchemaFields

-- GersonGalang - 23 Jan 2007

Edit | Attach | Print version | History: r6 < r5 < r4 < r3 < r2 | Backlinks | Raw View | More topic actions
Topic revision: r6 - 03 Dec 2007 - 01:15:23 - GersonGalang
 
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback