Computing at the Department of Mathematics
University of Houston
Goal: Provide necessary information about the computational resource.
Need help for problems related to computing and computers: send e-mail to 'help@math.uh.edu'.
(Please do not knock the door, show up in person and demand the help immediately.)
- The List of the Department Computing Servers: Use command "compute-status"
- compute-gpu1: 64 cores and 256GB RAM: Nvidia Tesla P100 16Gb
- compute-gpu2: 12 cores and 128GB RAM
- compute00-03 have 28 cores, 56 threads and 128GB RAM.
- compute04 have 40 cores, 80 threads and 512GB RAM (designated computing server).
- compute05-07 have 32 cores, 64 threads and 128GB RAM.
- compute08-11 have 10 cores, 20 threads and 48GB RAM.
- compute12-14 have 64 cores , 128 threads and 512GB RAM.
- Memory issue
- Always use "gzip filename" to compress your file and use "gunzip filename.gz"
to uncompress your file.
- Use the folder "servers/storage" under your home directory to store those files you do not use for now and those larger data files.
- If you need larger quota for your research, contact system managers via "help@math.uh.edu".
- The rules of running jobs on the Department Computing Servers:
- Only use the commands low-priority and high-priority to run your jobs: for example, use low-priority ./a.out to run your job a.out with the low priority.
- Each user can run at most two jobs with the command high-priority on each computing server, but
the total number of high-priority jobs should be limited to the number of computing servers,
which is 21 right now.
- The total number of low-priority jobs and high-priority jobs can be run on each computing server is up to the HALF of
the number of the cores. For example, on compute05, any user only can run at most two high-priority jobs and 4 low-priority jobs.
- Run simple jobs on your own desktop since many of you have a very fast multi-core desktop computer.
- Read and write into the local storage: There are two local directories for storage, '/scratch' and '/tmp', on each of the computing servers. You can create folders in these directories and read/write files in these folders to save the time from writing huge files in your home directory via the local low speed network. But please move these files back to your home directory once the job is done. The files stored at '/tmp' might be removed after certain times.
- Matlab users: To save some licenses of the Matlab for other users, please do run several Matlab jobs
on each computing server. Also please 'logout' from the Matlab toolboxes if you do not need them anymore.
- If any user violates the above rules, Dr. Pan will ask the user to terminate some of their jobs or
ask Jason to kill all jobs of this user. The user will get a warning for the violation at the first time. But for each repeated one,
the user will be get a suspension of the right of using the department computing resources for
one week.
- If anyone needs to run more jobs for some kinds of emergency, please talk to Dr. Pan to locate some other possible computing resources in order to avoid possible suspension.
- If you find any user abusing the computing resources, please send an e-mail to the Director of Computing: Dr. T.W. Pan at 'pan@math.uh.edu'. Dr. Pan will look into the situation.
- Locate available cores:
- Use 'compute-status' to find out how many cores are using right now among all computing servers. If load < cpus, there is some available computing power.
- Use 'compute-status -w' to find out how many cores are using by which users.
- After you log into a computing server, use 'top' to find out how many processes are running on the current compute server.
- Please do not expect that the fastest servers are always available.
- Some Tips for using these computing servers:
- Always use option '-O' when compiling your C codes and FORTRAN codes.
- Need help for problems related to computing and computers: send e-mail to 'help@math.uh.edu'.
(Please do not knock the door, show up in person and demand the help immediately.)
- Printing quota: Each student user
can print up to 300 pages per month. If you exceed the limit, you will not be
able to print anything, including quiz, exam, report and etc., for the rest of the month.
- Some useful Linux Commands:
- 'ssh': connect to these computing servers from your desktop,
for example, "ssh compute10" can connect to compute10 from your desktop via the 'terminal'.
- 'exit': terminate the connection established by the command 'ssh'.
- 'top': find out how many processes are running on the compute server.
Please do not submit any more 'higher-priority' jobs if all cores are using at the moment for the
'higher-priority' jobs.
- 'cat /proc/cpuinfo': find out what kind of the CPUs and its speed are on the computing server.
- 'user-search': find the user. E.g., 'user-search tsorng' gives user id, 'pan' and user name 'Tsorng-Whay Pan'.
- 'clean-firefox-locks': When firefox was not terminated cleanly, you just can not lunch it again.
This command cleans the firefox log and allow us to launch it.
- Some softwares:
- Here are commands for some of available FORTRAN and C compilers:
- pgf77: The Portland Group Inc. Fortran 77 compiler
- pgf90: The Portland Group Inc. Fortran 90/95 compiler
- pgf95: The Portland Group Inc. Fortran 90/95 compiler
- pghpf: he Portland Group Inc. high Performance Fortran compiler
- pgcc: The Portland Group Inc. ANSI and K&R C compiler
- pgCC & pgfc++: The Portland Group Inc. C++ compiler
- g++: GNU project C and C++ compiler
- gcc: GNU project C and C++ compiler
- g77: GNU project Fortran 77 compiler
- the above commands are located at '/usr/bin'. Use 'man' to find out their options, e.g., 'man pgf77' can give
manual of 'pgf77' with all available options on screen.
- Matlab (the statistics toolbox and symbolic toolbox are available): command is 'matlab' when executing on the 'Terminal' or launch it from the manual bar
- Mathematica: command is 'math' when executing on the 'Terminal' or launch it from the manual bar
Date of last change: March 26, 2019.