Registration Guide Register for HPC Access to HPCNew User Guide How to Run Batch Jobs
Portal User Guide
About the Service Subscribe to Pay-Per-Use Service Subscribe to Condominium Service Terms and ConditionsUser Guide
High Performance Workspace Service Utility Storage ServiceResearch Data Management ServiceCentre for Computational Science & Engineering
| Need help? | ||
![]() |
||
An Introductory Guide for New HPC Users
As a new user, when you registered a HPC account, the first thing you want to know may include one or more of the following:
- Which hosts can you login and what is the best ways to access the hosts remotely?
- How to transfer files between the hosts and your PC or another Unix system in your lab?
- After you login to one of the host, which directories can you use to run your jobs?
- What are the batch LSF queues for running long hours computing jobs?
This introductory guide is to answer those questions. Together with the relevant links provided, we hope that it will help you to get started smoothly.
At HPC, all systems are managed by the LSF (load sharing facility). With the LSF commands, you can submit your jobs to different hosts with different features, capabilities and limitations. You can monitor your jobs while they are running or check the load and status of each host to chose your submission queues and hosts. Therefore, before you start your work, do check the LSF Use Guide in HPC home page or read the man pages (type man followed by the command) for different commands to familiarize yourself with LSF commands and the queues and hosts used in the LSF environment. Please note that interactive jobs and background jobs submitted with ampersand (&) have a CPU limit of 30 minutes, they will be terminated after they reach this limit. The recommendation is to submit all compute intensive jobs as batch jobs through the LSF.
Some users are new to the Linux operating system. To learn more about Linux, you can read books on Linux; or simply type a "linux commands" in the internet search engines to search, you will get dozens of web pages with listing and explanation of Linux commands. A compiled list of the Linux commands is also available here for your reference, click here to download.
As a new users, you may have problems with our system or software from time to time, the HPC home page is at your finger tips all the time. Check different pages to find solutions for your problems.
Please send email to for problems and enquiries. A more specific description of your problems with error messages and so on will help us investigating and solving the problems.
Information on seminars, workshops, new hardware and software release, software upgrade, down time for system maintenance, etc., will be sent to you regularly through email. Please take time to read them in case your miss the important events and information.
Wishing you enjoy using HPC resources!
(1) HPC Account
HPC resources are open to all NUS staff and students who have a valid NUSNET id with the need to access high-performance computing resources.
- Click here to read the procedure and start registration.
- After registration, wait for about 1 hour, you will be able to use the userid and password to access all HPC resources. The NUSNET password is valid throughout all the hosts.
- As the access control and authentication of all HPC servers are through NUSNET account, change of NUSNET password will automatically take effect on all HPC servers.
- You can change you password on any HPC login hosts using "passwd" command. This will also change your NUSNET password.
(2) Accessing HPC Hosts
At HPC, we provides various kinds of compute servers and graphical workstations to meet the users' requirements for different hardware platforms and operating systems. The compute servers are used to run time-consuming compute intensive jobs, while the graphical workstations can be used to do the pre- and post-processing works and graphical display.
The follows are different ways to access these hosts:
- From any Unix/Linus system on campus:
- use ssh plus the hostname/ip address to access the following hosts/clusters:
-
Hostname Note atlas4-c01.nus.edu.sg HP Xeon Two Sockets Quad-Core 64-bit Linux cluster, RHEL 4 atlas4-c81.nus.edu.sg HP Xeon Four Sockets Quad-Core 64-bit Linux cluster, RHEL 4 atlas5-c01.nus.edu.sg HP Xeon two sockets Quad-Core 64-bit Linux cluster, RHEL 5 atlas6-c01.nus.edu.sg HP Xeon two sockets Hexa-Core 64-bit Linux cluster, CentOS 5 atlas7-c01.nus.edu.sg HP Xeon four sockets 10-Core and two sockets Hexa-Core 64-bit Linux cluster, CentOS 5 - From any PC using Internet Explorer or Firefox web browser
- http://nushpc.nus.edu.sg
Using SSH or PuTTY Client:- Install a SSH Client here or download PuTTY and install it on your PC and use them to access the HPC hosts.
To access the graphical workstations remotely:
MobileVisLab allows you to run graphical applications on the powerful workstations and display on your desktop.
(3) File Transfer
- From any UNIX/Linux system on campus,
- Secure FTP is recommended for file transfer, please check the availability of scp or sftp.
On windows system:
- User's home directory is mounted to U: drive when login to NUS domain. You can transfer files to your HPC home directory by drag and drop files in a Windows Explorer.
- If your computer is not joined to NUS domain, you can mount your HPC home directory (\\fs3\svu\username) from a Windows Explorer. Login to NUS-VPN is required when accessed outside of NUS network.
- Using the SSH secure File Transfer or SFTP in PuTTY.
- You can also access the HPC home directory for file transfer via the NUS HDrive online service (VPN not required)
(4) Working Directories
After login to any of above hosts, you will be able to access the following directories.
Directories Feature Disk Quota Backup Description /home/svu/$USERID Global 4 GB Snapshot Home directory, U: drive on your PC. The longest snapshot backup is 10 days. /hpctmp Local on All Atlas cluster 500GB Snapshot Working directory. Files un-touched for 60 days are automatically removed.
Details and instructions on how to use the working directory /hpctmp is available at the page of the High Performance Workspace for Computational Clusters.
- Type "quota -v" to check your disk quota for your home directory, use "df -k" command to check the free space left in a file system.
- Please do house-keeping work from time to time to make sure your disk quota is not exceeded, otherwise, you'll have problems to access HPC hosts and run jobs.
- More disk quota is granted if there is a need, submit the service request form for your request.
(5) Queues & Hosts for Submitting Jobs
Users are required to submit their time-consuming jobs to the compute servers through LSF batch queue system, instead of just submitting to the background with an ampersand (&) or running interactively in the foreground. Different hosts accept jobs from different batch queues. The following is a list and description of hosts and batch queues, for details, please check the LSF Guide on the Technique Information page.
Queue Names Execution Hosts CPU Limit Memory Limit Description linux64 All atlas clusters No 4096 MB Queue for jobs on all Xeon 64-bit systems (atlas 3, 4, 5, 6) atlas4_large atlas4 cluster No 20 GB Queue for large memory serial jobs on atlas4 cluster atlas4_parallel atlas4 cluster No 40 GB Queue for parallel jobs on atlas4 cluster atlas4_parallel_test atlas4 cluster 240 min. 40 GB Parallel test queue for atlas4 cluster atlas4_128 atlas4-c88 No 100GB Queue for very large memory jobs atlas5_large atlas5 cluster No 20 GB Queue for large memory serial jobs on atlas5 cluster atlas5_parallel_test atlas5 cluster 240 min. 32GB Parallel test queue for atlas5 cluster atlas5_parallel atlas5 cluster No 32GB Queue for parallel jobs on atlas5 cluster
atlas6_large atlas6 cluster No 72 GB Queue for large memory serial jobs on atlas6 cluster atlas6_parallel atlas6 cluster No 48 GB Queue for parallel jobs on atlas6 cluster atlas6_parallel_test atlas6 cluster 240 min. 48 GB Parallel test queue for atlas6 cluster atlas7_large atlas7 cluster No 200 GB Queue for large memory serial jobs on atlas7 cluster atlas7_parallel atlas7 cluster No 40 GB Queue for parallel jobs on atlas7 cluster atlas7_parallel_test atlas7 cluster 240 min. 40 GB Parallel test queue for atlas7 cluster openmp atlas5, atlas6, atlas7 clusters No 48GB Queue for shared memory jobs on atlas5, atlas6 clusters
g09_linda All atlas clusters No 16GB queue for running parallel g09 linda jobs short Cheetah 1440 min. 16 GB Queue for short running jobs which can finish within 24 hrs. short_parallel Cheetah 1440 min. 40 GB Queue for short running parallel jobs
(6) Linux Command
HPC hosts are all running Linux operating systems, being familiar with Linux command will surely help you in your work. There are many ways to learn more about Linux:
- Linux books from library
- You can find many books on Linux concepts and commands at the library.
- Web pages in the internet
- Search the internet with key words like "Linux command", you will get dozens of web pages listing the Linux commands.
- You can download the following command summary for your reference:
- Unix Commands.
(7) HPC Home Page
Remember to check the HPC home page for any information regarding HPC resources and services.
