It was highlighted in the previous issue that to make HPC resources and technologies more accessible to the Masses, the environment has to be made more user-friendly. This round, we will introduce some new services that will further enhance those important attributes. Starting from June 2008, researchers will be able to apply for HPC account online and access a wide range of HPC resources and services under one roof through the newly developed HPC Portal. In another related initiative, we will enable the use of the common account and password currently used for accessing NUSNET and email to access both the Windows and Linux/Unix based HPC systems. Researchers will also be able to access a common storage/file system from both Windows and Linux/Unix environment, which will enhance their ability to perform research computation anytime anywhere. In the area of application support, a new HPCBio portal will be launched in the 3rd quarter of this year to introduce a more intuitive way of accessing popular Life Sciences and Bioinformatics applications through the web interface. Besides improving the ease of access, there is also no lack of effort in advancing the capacity and capability of the central HPC resources. In the first four months of 2008, a monthly usage average of 153,873 CPU hours had been generated by the researchers. If one were to perform this amount of computation on a single desktop, it would have taken more than 6,411 days or more than 17 years to complete. More resources will be added in the next few months to push such computation capacity and capability to the next level. More details of such effort can be found in the article on "Getting More Work Done at a Faster Rate." Last but not least, don't forget to check out the new "Tips & Tricks" section where you can find ways to work more effectively in our HPC environment.

A new integrated portal for SVU's various resources is being implemented, and promises to provide users with greater ease of use and convenience. Leveraging on NUSNET Active Directory, and a common file system for all users that is accessible from both Unix and Windows, the NUS HPC Portal will enable users to access SVU's HPC resources faster and easier from anywhere that they may be. To find out more, read on.

With the Unix/Linux authentication integration with Windows Active Directory, SVU users will soon be using the same account and password to login to HPC academic and research computing resources (mainly Linux based), NUSNET and NUS Email. The integration can bring and offer users the following: 1) accessing to HPC resources with the same login name and password they use everyday to check their NUS email and to login to NUSNET; and there is no need to request for password reset for HPC resources any more; 2) new password is synchronized to all SVU systems immediately after they changed it, compared to at least 30 minutes lapse needed for the synchronization previously; 3) facilitating the ease of file sharing and accessing among Windows and HPC Unix/Linux environment; 4) providing single-sign-on login to HPC systems and more details will be given soon.

The HPCBIO Portal is a portal that provides NUS Life Sciences researchers convenient access to applications and compute resources. There are over 30 applications for Bioinformatics and Molecular Modelling. Each application has a web based graphical interface for input of application parameters and job parameters. A useful feature of the Portal is that the user does not need to know/find servers/queues to run his jobs. When he submits the job, the Portal finds a suitable server to run it. To find out more about the Portal, please read on.

With the expansion of the central HPC resources last year, researchers are now able to execute 3 times more jobs per month, with a much shorter queuing time and turnaround time. Further expansion scheduled for July this year will further enhance researchers' ability to solve larger-scale and more challenging computational problem. Please read on for more details ...

1, How to Find Idle Resources to Run More Jobs in Two Steps:

• Step 1, Issue command "lsload" at the command prompt, check the column "ut" and find the lowest loaded system and its hostname.
• Step 2, Issue command "bqueues -m hostname" to get the queue names through which you can submit your jobs to those lower loaded systems. Choose the serial queues or the parallel queues to submit your serial or parallel jobs accordingly. Please note these queues are not for job submissions: monitor, esvu and all the login queues.

2, How to Run Multithreaded Jobs through a Serial Queue

Submit your multithreaded jobs using the following command:

bsub -q linux64 -m multicore ./my_executable

The compute nodes on the atlas3 cluster have two dualcore CPUs which allow you run parallel jobs with 4 threads. This is suitable for those not very large parallel jobs. For example, Gaussian jobs submitted this way with a "nproc= 4" in the route section of the input file can get almost 4 times speedup.