Each account is linked to a MySQL user with the proper permissions. In order to log in the DIET WebBoard's end users interface, you must use your account name and your password.
Once it is done, you will have access to the main page.
This page is specific to each application. From here you can submit data to generate a "job" made of one or more "workunits".
For every application you can define a "name" for your job (Project name) and an "annotation" (Project annotation), which will help you to find your job in the job lists.
Also the "debug info" option is available for all applications, if checked, it will include standard and error output of the job, in "stdout.out" and "stderr.err" files, in the results packages.
This page will show the list of jobs you have submitted.
If you don't want to see a job in the list anymore, this list only displays "non-classified" jobs. You can force the page to display all your jobs by clicking on the link at the bottom.
The table shows many information about your jobs :
Name : the name you have given to the job, when submitting it.
Workunit done : the number of workunit processed by the grid. As checks are performed on the results to determine if no obvious error can be detected, it is possible that in certain cases this number decrease temporarily.
Workunit total : the number of workunits in the job.
First sent : the date when the first workunit was sent to DIET.
Last received : the date when the last received result of this job was registred.
State : the state of the workunit. It can be "running", "completed" , "suspended" or "failed".
ID : the unique identifier of the job in the DIET WebBoard.
When you click on the number in the "Workunit done" colum, you acess the Job checking page.
When you click on the ID you have acces to jobs results page.
On this page you have acces to all the parameters of the job.
If the job is completed you also have a link to download all its results.
When the number of workunit is not too high, you can see a list of all the results from the completed workunits of the job. When there are more than 100 workunits completed, you only have a link, clicking on it will force the interface to display all the results (but it may take some time).
This list shows the DIET WebBoard ID of the workunit, the device which made this result, the memory size of the result file, when the processing of the workunit started and finished, the state of the workunit (it can be : "waiting", "requested", "running", "completed", "failed", "suspended") and the time elapsed on the device by the workunit.
When the number of workunit is not too large, statistics graphs are generated :
You can see on the first graph which device group (cluster) have participated to the processing of the job. The second chart shows the repartition of the results file size, and the third chart represents the repartition of processing times.
Finally, when you do not want to see this job anymore in your job lists you can classify it. This will not delete the results, the data or the job, it will only delete temporary files (see Downloading section) and reduce the loading time of several pages. Of course if the job you are viewing is already classified the link will un-classify it.
From this page you can set the priority of your jobs between each other.
How do we decide which workunits are sent to DIET?
We make a list of all "waiting" workunits for the application.
We order the list by "priority" then by "random" (workunits with the same priority are in a random order).
We submit the workunits of the ordered list until the Max. workunits of the application is reached. We skip workunits of jobs which already have an amount of running workunits equals (or higher) than the job's max. workunits.
On this page you can set your jobs' max. workunit, and their priority.
For example you can decide that older jobs will always run first, to do that set an higher priority to old jobs.
Another example would be to share equally the ressources, if you have 5 jobs and 45 slots, you can set the maxworkunits to the 5 jobs to 9. disadvantage of doing that is that you will have to reallocate the 9 free slots when a job is finished. The default priority is 10, the default max. workunits is equal to the max. workunits of the application. This setting will be disadvantageful for small jobs (few workunits) as their probability of being choosen will be low. Small jobs should have dedicated slots, or an higher priority.
Results of the workunits are stored on storage spaces. These storages can be on the web server or on a remote machine.
Before downloading the results, these files must be copied on the web server.
However, when you request large files such as compilation of a job's result, the web server can take a long time to copy the results from the storage then generate an archive, and finally send it to the user : recent results of jobs are stored on a temporary storage space.
To manage this temporary space, a download manager page was made :
Note that job that are "classified" have their temporary file deleted and do not appear on the list any more.
From this page you can check your jobs and interact with them.
Fist you can see the state of the job (running, completed, suspended,...) and the number of workunit done.
If the job is running, you have a link to suspend the job. Of course you can also resume suspended jobs.
example :
You can also classify the job from this page (classified jobs will not appear anymore in the default job list, but their results and statistics will remain stored, and available). If the job you are checking is already classified, you can "un-classify" it.
If 10% or less of the results have been processed, you can delete the job. This will also delete the data, workunits, results, and statistics about this job.
The table represent the list of available results : you can see the ID of the corresponding workunit ("Workunit ID"), the state of the workunit ("state"), the ID of the result ("Result ID"), the file size of the result ("Result size", this is also a link to download the result), and when the result was received by the webboard ("Received", this is also a link to see the standard and error output of the result, available only when debug mode is activated when you submit the job).
The last column of the table can be used to select multiple results. Click on the button at the bottom of the page to cancel all these results and submit the workunits to DIET again.