Introduction for new users and students
This wiki is a collaborative initiative by the department to facilitate peer knowledge exchange. Users are encouraged to extend and update the information on this website.
New student or staff
Finding what you need
This wiki may serve as a reference for using certain softwares and computer systems. A few notes on usage:
- Use the search bar at the top of the page to query specific keywords.
- If a page doesn't exist, creating a new page and back-linking to other articles is easy.
- The preferred language is English but pages may also use Chinese. The markup language used by this site is called Wikitext.
In addition the wiki, the department runs a few other cloud services, such as:
- Our own Gitlab, for all staff/students
- Our high performance computing cluster.
- You can also use this for storage, but staff/students typically already have a large GoogleDrive or OneDrive.
- An Rstudio Server (Experimental)
- A research PACS server (Under development)
New HPC user
What and Why
There are many websites introducing and explaining the concepts of high performance computing. To quote a page from Iowa state university:
"An HPC cluster is a collection of many separate servers (computers), called nodes, which are connected via a fast interconnect."
It may not be necessary to learn everything about using the hpc cluster, here's some use-cases:
- You can run code written in Python, MATLAB, Julia, Perl, C, C++
- You can store/backup data
- You can work in a remote desktop (Xfce by default)
- You can run a webserver, e.g. a Jupyter Notebook
Using our cluster
If you wish to use our hpc cluster, please create an account. If you are already familiar with working on a remote Linux system you probably only need to consult the GPU cluster page. Most uses will require you know how to connect to an interactive shell session over SSH. See our guide on shell basics.
Professionally managed 'university tier' clusters typically have more nodes and employ a more complex usage model, involving job scheduling, additional node types, and stricter access policies. In comparison, our hpc cluster has lower access threshold and relies on user discussion to resolve resource conflicts. Write access is typically restricted to non-critical folders, so don't be afraid to break anything.
Installations can be categorized into language-specific packages and general Linux packages.
Users installing python modules can use
pip, preference in that order. If you install a package with pip, it will be invisible to conda. In order to avoid package conflicts, users are advised to use virtual environments.
Users following generic installation instructions typically find the package manager
yum doesn't work. To use new software, users have full permissions to install binaries manually into their home folder. Usually binaries can be downloaded directly or compiled. Alternatively,
conda may provide the package. In some cases new software requires many additional dependencies, although manual install is still possible, in this case it is suggested to request an admin to install the package.
Please refer to our main department website at radiology.hku.hk for contact details.