Difference between revisions of "Common-accounting-model ABANDONED"

From Gcube Wiki
Jump to: navigation, search
(Service)
(Execution)
Line 19: Line 19:
  
 
=== Execution ===
 
=== Execution ===
Regarding the Execution resource type, there are two sub-types, according to the PE2ng's structure which is composed by two main layers. There is the Workflow layer that is more abstract, constructing workflow plans, supporting various adaptors and is aware of jobs as a whole. There is also the Execution layer, also a Service, where the actual execution takes place and is aware of more detailed stuff.
+
This specification will be used to take into account information about services running jobs on the infrastructure (Workflow Engine, Execution Engine, Statistical Manager, Aquamaps).
  
Discriminating those layers:
+
For this resource type, there are two sub-types:
* Workflow layer is aware of:
+
Number of jobs submitted and adaptor that were used
+
Execution nodes that will be used (scale out) per job
+
  
* Execution layer is aware of:
+
==== Job ====
Statuses of execution jobs (success/fail)
+
Contains the information about the overall job, that will be partitioned in N Tasks.
also GHN hosting node information of every execution node is available to Workflow, harvested through Registry, containing info such as location, cpu load (week, day, hour,...), memory, disk space etc.
+
  
 +
Specific Job properties:
  
==== Plan ====
 
 
Specific Plan properties:
 
 
* adaptorInUse : adaptor in use for the job (e.g. search, data-transformation, etc.)
 
* vmsUsed : number of the VMs used by the job.
 
 
* jobId : an unique identifier for the job
 
* jobId : an unique identifier for the job
 +
* jobQualifier : qualifies the job in terms of algorithm type or job type (e.g. search, data-transformation, etc)
 
* jobName : name of the job
 
* jobName : name of the job
 
* jobStart : the instant the job start running
 
* jobStart : the instant the job start running
 
* jobEnd : the instant the job ends its execution
 
* jobEnd : the instant the job ends its execution
 
* jobStatus: completed/failed
 
* jobStatus: completed/failed
 +
* vmsUsed : number of the VMs (gHNs) used by the job.
 
* wallDuration : duration between the instant the job start running and the instant the job ends its execution.
 
* wallDuration : duration between the instant the job start running and the instant the job ends its execution.
* cores : number of available cores per job.
 
* processors : number of available processors per job.
 
  
==== Execution Engine ====
 
  
Specific Execution Engine properties:  
+
==== Task ====
 +
Contains the information about one slice of the overall Job.
 +
 
 +
Specific Task properties:  
  
* refHost : hostname of the vm
+
* jobId : reference to the Job that generated this Task
* refVM : Execution Engine resource id or gHN id
+
* refHost : hostname of the virtual machine (gHN)
* usageStart : the earlier usage time of the Execution Engine
+
* refVM : virtual machine id (gHN)
* usageEnd: the latest usage time of the Execution Engine
+
* usageStart : the earlier usage time of the Task
 +
* usageEnd: the latest usage time of the Task
 
* usagePhase: completed/failed
 
* usagePhase: completed/failed
* inputFilesNumber : number of input files to the Execution Engine
+
* inputFilesNumber : number of input files to the Task
* inputFilesSize : dimension of input files to the Execution Engine
+
* inputFilesSize : dimension of input files to the Task
* outputFilesNumber : number of output files from the Execution Engine
+
* outputFilesNumber : number of output files from the Task
* outputFilesSize : dimension of output files from the Execution Engine
+
* outputFilesSize : dimension of output files from the Task
* overallNetworkIn : overhead of the input traffic over the network to the Execution Engine
+
* overallNetworkIn : overhead of the input traffic over the network to the Task
* overallNetworkOut : overhead of the output traffic over the network from the Execution Engine
+
* overallNetworkOut : overhead of the output traffic over the network from the Task
 +
* cores : number of cores per Task.
 +
* processors : number of processors per Task.
  
 
=== Service ===
 
=== Service ===

Revision as of 13:32, 24 May 2013

Scope

This library contains the definition of the resource accounting record.

Data-model

The structure of a generic accounting record (Usage Record, UR) will be composed of a set of common fields for all resource types, in particular:

  • id : an unique identifier for the UR
  • consumerId : the user actually consuming the resource (optional, for future purposes)
  • createTime : when the UR was created
  • startTime, endTime : the time window the UR refers to
  • resourceType : the type of resource the UR tracks
  • scope : the scope of the resource
  • resourceOwner : who owns the resource and/or who creates the UR

Furthermore, for each UR there will be a section to be filled with the specific properties per resource type (key-value pairs).

Resource Types

The resource types we've identified are: Execution, Service, Data-access and Storage.

Execution

This specification will be used to take into account information about services running jobs on the infrastructure (Workflow Engine, Execution Engine, Statistical Manager, Aquamaps).

For this resource type, there are two sub-types:

Job

Contains the information about the overall job, that will be partitioned in N Tasks.

Specific Job properties:

  • jobId : an unique identifier for the job
  • jobQualifier : qualifies the job in terms of algorithm type or job type (e.g. search, data-transformation, etc)
  • jobName : name of the job
  • jobStart : the instant the job start running
  • jobEnd : the instant the job ends its execution
  • jobStatus: completed/failed
  • vmsUsed : number of the VMs (gHNs) used by the job.
  • wallDuration : duration between the instant the job start running and the instant the job ends its execution.


Task

Contains the information about one slice of the overall Job.

Specific Task properties:

  • jobId : reference to the Job that generated this Task
  • refHost : hostname of the virtual machine (gHN)
  • refVM : virtual machine id (gHN)
  • usageStart : the earlier usage time of the Task
  • usageEnd: the latest usage time of the Task
  • usagePhase: completed/failed
  • inputFilesNumber : number of input files to the Task
  • inputFilesSize : dimension of input files to the Task
  • outputFilesNumber : number of output files from the Task
  • outputFilesSize : dimension of output files from the Task
  • overallNetworkIn : overhead of the input traffic over the network to the Task
  • overallNetworkOut : overhead of the output traffic over the network from the Task
  • cores : number of cores per Task.
  • processors : number of processors per Task.

Service

This specification will be used to take into account information about the services invocations.

Specific service attributes

  • callerIP : IP address that originated the service call
  • invocationCount : number of invocations (aggregated information)
  • averageInvocationTime : average invocation time (aggregated information)
  • serviceClass : name of the service class
  • serviceName : name of the service

Data-access

Specific Data-access properties:

  • sourceId: the identifier of the Tree Manager source which is the target of a read/write operation
  • operation : the name of the read/write operation performed via the Tree Manager over a given source
  • treeId : the identfier of a tree within the data source which is the target of a given read/write operation performed via the Tree Manager
  • treeCount : the number of trees within the data source which are accessed/written as the result of a given read/write operation performed via the Tree Manager

Storage

This model specification will be used to take into account storage resources and timeseries/services using DBs backend.

Specific storage attributes:

  • operationType : GET, PUT, UPDATE, DELETE
  • targetResource : URI representing the storage resource
  • fileDimension : dimension of the storage resource
  • hostname: hostname of the host where the storage library is invoked