Grid Gateway Interface (GGI) is a simple web-based interface to Grid Engine distributed computing management system (for details about Grid Engine see gridengine Project home).
GGI is used to submit computationally-heavy tasks directly from the web browser query input form into our high performance computing cluster. Queries are queued and then run as resources permit. Users are able to track the progress of their queries and retrieve results at a later time. It is also possible to submit large number of queries at once (current limit is 100) without having to wait for each one to complete.
This Guide will walk you through the Grid Gateway Interface window and explain how to manage your queries.
Some terminology used:
When you push Submit Query button on a query form page, you are transferred to GGI web page (you can also get there without submitting a new query by means of Check Status button). At the top of the page, below the Grid Gateway Interface header with the logo and documentation links, you will see a few lines of information listing the name of the service you are using, your session ID, and overall status of the grid backend.
We will discuss Session ID later. Service Name links back to the particular web page you have used to submit your query. There are several web services installed on our server that utilize the same GGI system; by checking the Service Name you should always know which one you are currently using.
Grid Status section lists a few useful hints about the current status of the computational resources available: Load indicates how heavily grid computers are currently utilized, Health reflects the percentage of perfectly functional nodes on the grid, while Jobs lists the total numbers of all Pending and Running jobs. Overall, this information allows you to get some estimate about how long your query may take to complete. There is no exact rule however, since jobs may have very different resources requirements. If you notice that Health indicator is close to 0% and stays that way for a prolonged time, you are welcome to kindly remind us that our grid is suffering!
Section below Query Status line is where you will find information about all of your jobs listed, together with the links and checkboxes to control them. There are two parts: top one is about Completed jobs while Pending and Running jobs are listed below. When you submit your query, it will always go to Pending list first. You can check its status at any time by clicking on Refresh Status button. Note, that GGI page will not refresh automatically, you should use Refresh Status button to obtain most current status information. Please, do not abuse server by excessive refreshes. Keep in mind, that internally the status data is updated every 5 seconds. The time when last report was generated is listed at the bottom of the page. This corresponds to server's local time (US/Eastern). Depending on your location, time zone and national daylight saving regulations, this can be quite different from what your clock shows. Here is what each column in the Pendig/Running section means:
qw for queued, awaiting
execution; t for being transferred to one of the
grid's nodes for execution; or r for running jobs.
There are other state codes but you should not normally encounter any of those.
For instance, when there is E included in the
status code (e.g. Eqw) it means that the job
caused internal Grid Engine error and cannot be executed. You may try to
resubmit such job and if error persists - send us a bug report (do
not forget to include job's numeric ID)After your query finishes, the results are listed under Completed heading of the Query Status section. There are two new columns appearing in this listing, with the links to the files produced:
Note, that the View link under Errors only shows up if there are some exceptions caught during the execution of the analysis pipeline. This should rarely happen and usually indicates either a bug in the software or a problem with the grid configuration. If reproducible, you may want to report it so we can get it fixed.
On the other hand, contents of Results column may also indicate some errors, often caused by incorrect values in user input. You should be able to fix such errors yourself by returning to the original query page and re-entering corrected values of parameters in a query form, then resubmitting the query.
Our Grid Gateway Interface server uses transparent login and user's session registration based on automatically generated session ID, a 40-character hash string that uniquely identifies user's account and computer from which a request has been made. There is no need to explicitly register on the server to use it. However, you should be aware that this system does not provide any reasonable level of security because anyone who gets access to your session ID will be able to login to your session and read and manipulate your data stored on the GGI server.
Session ID is kept in your browser's cookies so that you do not need to manually enter it every time you return to our server. Cookies are associated with the user's profile usually connected to a user's account on a particular computer. If you wish to access your GGI session from another computer and/or user account than the one you have originally used to start your session, then you will need to enter your session ID manually in the Session ID input field at the top of GGI page. After that, when you press Refresh Status button you will be redirected to your original session account. The session ID you have entered manually will be effective for the duration of your browser run; when you close and restart the browser it will revert to whatever default session stored in its cookies (or create a new one). You can however overwrite browser's default session ID with the manually entered one if you check the Overwrite default box located to the right of the Session ID input field.
Your GGI session ID, stored in your browser's cookies, is set to expire after 3 months of inactivity. All results of your analyzes will be also removed from the server after 3 months period since they have been obtained. Please, be aware that although we would try to keep your data for 3 months, we cannot guarantee this. It is possible, that in an event of unexpected emergency we will be forced to remove all users' data from the server. Hence, you are advised to extract the results of your queries as soon as they are available and save them on your local computer.
Please feel free to send us your questions, comments, suggestions for improving our services and bug reports. You may also wish to check GGI FAQ for answers to common questions.
| E-mail us | Last modified: 02/24/2008 |