This feature is part of the AXIS GridLink Module.
Please note that this feature requires the AXIS GridLink Module Licence.
Please refer to the help topic "AXIS GridLink Module" for the overview.
If an AXIS user has a batch job that he/she wants to run on a server farm, he/she can simply right-click on the batch from the list and select "Run as Remote Job" from the pop-up menu. A request is send to the server machine where AXIS Grid Link Controller is running and the dataset is closed. It is also possible to select more than one batches and submit them with one command. If the job is submitted successfully, the user can choose whether to monitor the batch status or not.
The user also can submit batch jobs to the controller without having to open the dataset. This is especially useful for people who use a slow network connection and have datasets located on remote machines. This is accessible from the menu "Tools - GridLink Module - Submit Remote Jobs". The user needs to specify a dataset, then selects batches from a list. If the version of the dataset is older than the version of the current AXIS, AXIS will ask the user whether to proceed. If the user chooses to continue, the dataset will be converted when being opened by the remote master.
If there are more than one controller registered in the System Parameters, the user will be asked to select one from the list. The controllers on the farm will automatically decide the number of helpers to be launched based on the availability of all CPU cores on the farm. In the case that there are more than one distributable jobs running on the farm at the same time, the farm will be split automatically based on each job's resource priority that you can specify here. A job with higher resource priority will receive more helpers. If the user doesn't need any helper to help this job, he/she can simply specify the priority as "No Helpers".
When an "Urgent" job is submitted, even if there’s another distributable job (non-urgent) is running in the same job queue, the urgent job will start right away as long as the master server has multiple CPU cores and multiple jobs are permitted to run simultaneously on the farm. Once started, the urgent job will take away all helpers that were helping the other job. The non-urgent job will run by the master alone until the urgent job is done.
Helpers launched for a job will be automatically closed once the job is finished.
In the event that some helpers failed, controllers can automatically start new
helpers on the farm.
In each job queue GridLink can run one distributable and multiple non-distributable jobs in parallel - up to 4 job in total but no more than the number of CPU cores on the master server.
While this feature may significantly increase farm throughput in the case of multiple long non-distributable jobs (e.g. DataLink macros) - in some cases it is desirable that jobs are executed sequentially one-by-one.
If the farm is busy, the controller will keep the batch job in a queue. By default, as soon as the farm becomes available, a remote AXIS master and helpers will be started to run this job. However, the user can specify that a job should not start until a specific file exists, or until a specific time.
Once the job is done, if the user is still monitoring this job since the submission, the monitor dialog will be closed automatically, the dataset will be reopened on the user's local machine and System Log messages ever logged during the run will be displayed, as if this job was run locally.
Before the user can submit a batch job to the server farm, all of the following conditions must be met on the remote AXIS master machine:
1) AXIS GridLink Controller is running and properly configured. (Please refer to help topic "AXIS GridLink Controller" for more details.)
2) a correct version of AXIS has been installed.
3) the user has both read/write permission on the dataset path and the Import/Export database.
The name of the remote AXIS master machine and the TCP port through which AXIS GridLink Controller is connected also need to be specified in the System Parameters. Please refer to help topic "Distributed Processing Tab in System Parameters" for more details.
Submit a Remote Job from Command Line
The user can also submit a remote job to GridLink from command line by using
the following 3 new options:
-GLC: the controller name
-GLT: the TCP port number. If not specified, the setting in the system
parameters will be used
-GLR: the resource priority. Can be "high", "normal" (default), "low" or "nohelpers"
-GLWaitFile: a file name. The job will not start until a file is detected
-GLWaitTime: a time in the format of "dd-mm-yyyy hh:mm:ss". The job will not start until this specific time.
The following are some examples:
AXIS -D:"C:\Datasets\Sample" -B:1 -GLC:mycontroller -GLR:high
AXIS -D:"C:\Datasets\Sample" -B:1 -GLC:mycontroller -GLWaitFile:C:\Spare\marker.txt
AXIS -D:"C:\Datasets\Sample" -B:1 -GLC:mycontroller -GLWaitTime:"23-05-2008 12:00:00"
For information about monitoring the status of remote jobs, please refer to the help topic "Monitor Remote Batch Jobs".