Contents
This FAQ provides general information for researchers who want to know more about the CÉCI and who contemplate creating an account. If you are looking for more technical documentation, please click on the "Documentation" tab.
1. About the CÉCI Get to know more about the CÉCI.
1.1 What is CÉCI?
1.2 What is supercomputing ? What is a cluster ?
1.3 Who can claim access to the CÉCI supercomputing infrastructure?
1.4 How are computing resources used and shared ?
1.5 How fast are the clusters ?
1.6 Is CÉCI related to PRACE ?
1.7 How can I access bigger (Tier-1, Tier-0) clusters ?
1.8 How can I influence decision making ?
1.9 May I store sensitive/personal data on the cluster ?
2. First steps and Quick start Discover how to get started using the CÉCI supercomputers.
2.1 How do I create a CÉCI account?
2.2 How do I connect to the clusters ?
2.3 How do I start working ? (Copying files, installing software, submitting jobs)
2.4 How do I acknowledge the use of CÉCI clusters ?
2.5 What are the rules for using the CÉCI clusters ?
2.6 Which cluster should I use ?
2.7 Which software is available ?
3. About my account. Have questions about your CÉCI account ? Keep reading.
3.1. Will my account expire at some point ?
3.2. My account has expired, what do I need to do?
3.3. When my account expires, do I lose all my data ?
3.4. Can I share my account with other colleagues ?
3.5. Can I share my data/code with other colleagues ?
3.6. I lost my private key! Can you send it to me again ?
3.7. I don't remember my passphrase! Can you remind me?
3.8. My email address changed. What can I do ?
3.9. Can I have more diskspace | more priority | sudo/root privileges ?
3.10. Can I renew my account from abroad ?
4. If this FAQ does not help...
4.1. I did not find an answer to my question here. What do I do ?
4.2. I still cannot find an answer.. Please help !
4.3 How can I make sure I get help quicky?
1. About the CÉCI
1.1 What is CÉCI?
CÉCI stands for 'Consortium des Équipements de Calcul Intensif'. CÉCI is thus a consortium which gathers together all supercomputing centers from the universities in Fédération Wallonie-Bruxelles, namely Université libre de Bruxelles, Université catholique de Louvain, Université de Liège, Université de Mons, and Université de Namur. The CÉCI is supported by the F.R.S-FNRS and the Walloon Region. See more information in the 'About' section of this website.
1.2 What is supercomputing ? What is a cluster ?
Supercomputing is a generic term associated with the fact of using a supercomputer. A supercomputer is to a computer what a supercar is to a car ; it does basically the same, except is does it much faster and all its components are designed for performance. A computer cluster is a particular type of supercomputer. It is a set of computers set up to be seen as one single, very large computer.
1.3 Who can claim access to the CÉCI supercomputing infrastructure?
Each University of the CÉCI consortium (ULiège, UCLouvain, ULB, UNamur and UMons) is in charge of accrediting its members for CÉCI cluster access. The official CÉCI consortium agreement, which governs the access and use of the clusters, is available here. See more specifically paragraph 6.4.
In practice, in order to ask for the creation of a CÉCI account, one needs to be officially affiliated with a university member of the CÉCI consortium, be holder of an official university email address, and be endorsed by a supervisor with a permanent academic or scientific position. If you do not have an official university address, please engage administrative procedures to get one.
In the case that obtaining an official university email address is too cumbersome with respect to the duration and/or nature of the collaboration, the holder of a CÉCI account who holds a permanent position within his/her university has the ability to invite a guest for a limited period of time. The procedure is slightly different than for a regular user. See FAQ 2.1 for the procedure to follow.
Please also see Question 2.5 What are the rules for using the CÉCI clusters ?.
1.4 How are computing resources used and shared ?
The computing power is distributed evenly across researchers and organized by a job scheduler to ensure fair access for everyone. Users request resources by submitting jobs to the job scheduler, which allocates resources to the jobs based on job priorities. The job priorities are defined according to past usage, current waiting time, and other more technical criteria as explained in this document.
1.5 How fast are the clusters ?
The clusters consist in many interconnected computing units so how fast it is depends on your program. If your job consists in a single serial program, the cluster will not be of much help ; it might even run longer on the cluster than on your personal computer. This sometimes is a misconception for new users: the power lies in the number of computing units, the available memory, and the fast network, but not in the individual power of one computing unit. The clusters only develop their potential when running parallel programs or when running several serial programs at the same time.
1.6 Is CÉCI related to PRACE ?
PRACE is the 'Partnership for advanced computing in Europe'. The mission of PRACE is to 'enable high impact scientific discovery and engineering research and development across all disciplines to enhance European competitiveness for the benefit of society.' Belgium is part of PRACE, and the CÉCI offers Tier-2 level supercomputers to help researchers develop their code and debug their jobs to apply for computing time on country-level Tier-1 clusters and Europe-level Tier-0 infrastructures.
1.7 How can I access bigger (Tier-1, Tier-0) clusters ?
The CÉCI also enables the users with access to the Tier-1 (national-level) and Tier-0 (European-leve) infrastructures.
The Walloon supercomputer Lucia is hosted at Cenaero. Access to Lucia is granted solely through 'projects'. A project must be created by a CÉCI user who holds a permanent position (e.g. Professor, etc.) within a member university. Other CÉCI users can then register to that project and will have access to the resources on Lucia (CPU time, disk space, etc.) that are allocated to the project. See more information about the project submission process on the Tier1 page.
See also:
The European supercomputer LUMI is hosted in Finland. Access to Lumi can be obtained by responding to calls issued either by EuroCC Belgium or EuroHPC.
More information:
1.8 How can I influence decision making ?
The CÉCI bureau gathers a few times per year. The easiest way to get messages passed along is to contact your local representative and communicate with them. They are all listed in the About page.
Another option is to respond to the survey that is organised regularly, which contains a free-comment form where you can express opinions and remarks. Surveys are announced on the CÉCI mailing list.
1.9 May I store sensitive/personal data on the cluster ?
The CÉCI servers are professionally managed and offer all the security guarantees according to current good practices security-wise, guaranteeing the absence of access by unauthorized third parties. But the clusters are multi-user machines, and protecting one's data from other users is the user's responsibility, it is not a built-in feature of the clusters.
More generaly, the question of good practices related to processing sensitive data is more of an administrative question than a technical one; the opinion of sysadmins is not relevant regarding the measures to be taken to store that kind of data, but sysadmins can help to put in place the measures that must be taken. These measures are normally enacted by the project's Data Management Plan if it exists, or failing that, by the project manager or the university's Data Protection Officer (DPO).
Furthermore, each system administrator has full access to the data on his cluster, and, by default, data are not considered to be sensitive. But the system admins are not necessarily member of the university that owns the data, and therefore are not included in any non-disclosure agreements, subject agreements for processing, etc. Remember that the CÉCI is made up of 5 administratively distinct entities.
In general, the Data Protection Officer's recommendations for data processing on third-party infrastructures will be either:
- to abstain, and only work on 'local' hardware ; or
- to encrypt the data, either
- in transit -- this is always the case on clusters with SSH
- at rest on disks (when not processed by a job) -- it is the user's responsibility to do so, and system administrators can help set up what is needed
- at work in RAM (for the duration of the job) -- this is almost impossible to ensure on clusters ; or
- to anonymize the data (make it completely impersonal even by cross-referencing information) -- anonymized data is no longer personal in nature but it is necessary to ensure that this is compatible with the processing carried out (generally ok for statistical analyses for example), and for this it is better to ask for help from statistics specialists ; or,
- to pseudonymize the data (replace personal elements with neutral identifiers) -- but this does not remove their personal character and in any case the "translation" key must not be stored on the clusters.
In summary, determining the best practices to follow is a step to be taken by the user with the help of the Data Protection Officer, Data Management Officer, etc. while system administrators can take care of installing software and help setting up the workflows, but if administrator rights are necessary for the implementation of the solution, this will almost certainly not be compatible with use on a multi-user machine of another university.
2. First steps & Quick start
2.1 How do I create a CÉCI account?
Please note that in order to create an account, you must be connected, either directly (Wifi, cable), or indirectly (via SSH or VPN) to your university network.
- If you are member of a university part of CÉCI :
The first step is to visit login.ceci-hpc.be and
enter your email address. From then on, your email address will be associated
uniquely with your CÉCI login. You will then shortly receive an email with a
web link. Click on that link; you are brought to a web form where you need to
enter your identification information. Your information is then reviewed by a
system administrator from your university. Once your details are confirmed,
your account will be created. As soon as it is created, you receive your
private key to the clusters by email (an attached file named id_rsa.ceci
.)
See also: Create a CÉCI account step by step tutorial.
- If you want to invite a guest :
You can create an account for your collaborators who do not have an email from one of the five universities members of the CÉCI. You must have a permanent position within your university and be a registered CÉCI user. If you are not, please register yourself first as in previous paragraph. By using this functionality, you bear the responsibility of the account created for your guest. Go to login.ceci-hpc.be/#invite_guest, and enter your email address and the email address of your guest (we need the official institutional email address of the guest. Addresses ending in @gmail.com, for instance, or other generic email addresses are not allowed). You will have to fill in the form sent to your email address, and your guest will have to send an SSH public key to the system administrator of your university. Detailed instructions will be sent by email to your guest directly.
2.2. How do I connect to the clusters ?
The clusters can only be accessed from a computer connected to a university network, either directly (cable or WiFi), through a gateway or through VPN. They must be accessed through a secure shell (SSH).
Operating systems based on UNIX (e.g. Linux, MacOS, Solaris, ... ) have SSH
programs readily installed that can be invoked with the ssh
command.
Windows, by contrast, does not, so you will need to install one.
Please, refer to the following information to learn how to connect to the clusters:
- Tutorial: Connecting from Windows OS
- Tutorial: Connecting from UNIX/Linux or MacOS
- CÉCI/CISM training sessions about SSH
Note that as the clusters are only available from the university networks, you must take extra steps to connect from your home or when you are abroad. Detailed information about this is given in the tutorials above, please make sure to follow them until the end.
2.3. How do I start working? (Copying files, installing software, submitting jobs)
The typical steps involve transferring files, data and code, possibly installing sotware, and finally submitting jobs.
The first step will most often be to transfer files (data and/or code) into your
home
directory on the cluster, a directory where you can store your personal
data and code, and which is accessible from all compute nodes of the cluster.
- Tutorial: Transferring files from/to a Windows computer
- Tutorial: Transferring files from/to a Linux or Mac computer
Note that each cluster has its own hierarchy of storage solutions but they all share a common storage visible from all of them.
The second step, is to check if the software you need is available, and if not, to install it.
- Reference: Using pre-installed software
- Reference: Compiling and installing software from source
Then, you will need to submit a job, i.e. write a submission script describing the resources you need and the operations you want to perform, and submit that script to the resource manager/job scheduler. The one installed on the CÉCI clusters is named Slurm.
- Slurm tutorial and quick start
- Slurm Frequently Asked Questions
- Slurm documentation
- Slurm Submission script wizard
All of this can be learned by attending the CÉCI/CISM training sessions.
2.4 How do I acknowledge the use of CÉCI clusters ?
In your scientific publications, please acknowledge the use of the CÉCI clusters, as well as the support from the local system administrators. Some universities have their own rules. Specifically:
The acknowledgement must include at least the following:
"Computational resources have been provided by the Consortium des Équipements de Calcul Intensif (CÉCI), funded by the Fonds de la Recherche Scientifique de Belgique (F.R.S.-FNRS) under Grant No. 2.5020.11 and by the Walloon Region"
Furthermore, if you used the Tier-1 cluster, it must be acknowledged specifically, as described here.
For publications written in French, the following can be used:
"Les moyens de calcul ont été fournis par le Consortium des Équipements de Calcul Intensif (CÉCI), financé par le Fonds de la Recherche Scientifique de Belgique (F.R.S.-FNRS) sous la convention n° 2.5020.11 et par la Région Wallone"
2.5 What are the rules for using a CÉCI cluster ?
Firstly, you need to adhere to the general rules related to the use of computers and networks at the different universities hosting the clusters you are using. They can be consulted at the following web pages.
- UCL: https://uclouvain.be/fr/decouvrir/acces-informatiques.html
- ULB: http://www.ulb.ac.be/ulb/greffe/documents/reseau.html
- ULiège: https://www.campus.uliege.be/cms/c_9109523/fr/modalites-et-deontologie
- UMons: http://portail.umons.ac.be/.../Charte de l'utilisateur des ressources informatiques.pdf
- UNamur: http://www.fundp.ac.be/organes/codeo/principes
Note that access to the Tier-1 facility hosted at Cenaero is subject to the additional terms and conditions described here.
Secondly, you need to conform to the 'good practices' described in the Fair usage of the CÉCI clusters document, and those prescribed in the Choosing and using a cluster wisely document.
Thirdly, make sure you acknowledge the use of the CÉCI clusters in your publications (See Question 2.4)
Fourthly, you must agree to be registered to the CÉCI mailing list (a dozen emails per year related to interruptions of services and CÉCI events).
Finally, you need to make sure to be aware of the conditions defined by your local HPC facility to access the clusters:
- UCL: http://www.cism.ucl.ac.be/login
- ULB: https://cc.ulb.ac.be/hpc
- ULiège: https://www.campus.uliege.be/nic5
- UMons: http://dragon1.umons.ac.be/dokuwiki/doku.php
- UNamur: https://www.ptci.unamur.be/wiki/doku.php?id=doc:index
2.6 Which cluster should I use ?
The choice of the cluster should firstly be based on the hardware the cluster offers. For instance, large parallel jobs should be run on a cluster with a fast network, and jobs producing a lot of data should be run on clusters with a fast file system. Jobs needing none of the above resources should run on clusters which do not have them so as to leave the resources available for the jobs depending on them. See more details and hints in the Choosing a cluster wisely document.
Nevertheless, sometimes, the availability of a commercial software will restrict the choice with the cluster, but free and/or open-source software can be installed anywhere, either by yourself in your home directory using configuration scripts when available, or by the system administrator. Do not let the availability or non-availability of such easy-to-install software guide your choice. Similarly, queue policies may limit your choice. For instance, time limitations are different from one cluster to another. Keep in mind, though, that well-written scientific software often implement checkpointing, a mechanism that allows stopping a computation moment and restart it at a later time.
Finally, exceptionally, the current load of the clusters can be used to choose a cluster to launch short jobs, say less than a day, to avoid monopolizing a resource which is underused.
2.7. Which software is available ?
On most clusters, several C/C++ and Fortran compilers are installed, often several versions of them are available too. Interpreters for popular interpreted languages (Octave, Python, R) are also often installed.
Many other software are also installed, mainly upon request by the users. See mode information here.
You can always ask the cluster system administrators to install software for you, or install software by yourself.
3. About my account.
3.1. Will my account expire at some point ?
You need to renew your account every year; it will otherwise expire. Don't worry, you will be notified by email one month prior to expiration so that you can renew it without interruption of service. Note that you still can renew your account once it has expired. To renew your account, simply go to the account creation page, and enter your email address (the same one you used to created your account in the first place), or click on the direct link you received by email. The procedure is similar to the account creation procedure.
3.2. My account has expired, what do I need to do?
If you still have the email notification that your account was about to expire, just follow the link it refers to. If you lose that email, go to login.ceci-hpc.be/#renew_account and enter the same email address you used to create your account in the first place. You will then proceed in a similar way than when you created your account.
Note that the process of renewing your credentials on all CÉCI clusters takes some time, but normally no more than 20 to 30 minutes. During that time window, it is advisable to load both the previous key and the new one in your SSH agent when connecting to the clusters. If you do not use an SSH agent, you will need to try both keys one at a time.
3.3. When my account expires, do I lose all my data ?
When an account expires, it is simply disabled. Files are left untouched and you will find them back when you renew your account and reconnect.
HOWEVER, please note the important two points hereafter.
Remember that files stored on scratch filesystems are subject to deletion during maintenance periods, or when the filesystem is full, whatever the status of the account to which they belong.
Files belonging to accounts that have remained in the expired status for a long time will be reviewed for permanent deletion on a per case basis by the system administrators from your university/research center no sooner than 500 days after the expiration date, and in agreement with the person you mentioned as 'supervisor' in the account creation form.
3.4. Can I share my account with other colleagues ?
No you are not allowed to do so. This is a very bad idea that would deforce all the security systems that were implemented to make the clusters and your data safe.
3.5 Can I share my data/code with other colleagues ?
Yes, definitely. It just is a matter of issuing the correct commands. See more information here
3.6. I lost my private key! Can you send it to me again ?
No, we do not keep the users' private keys. In the case you lose it, go to login.ceci-hpc.be/#retrieve_key, and submit your email address. Follow then the procedure as you did when you created your account in the first place.
3.7. I don't remember my passphrase! Can you remind me?
No, we do not keep the users' passphrases. If you can't remember it, go to login.ceci-hpc.be/#change_passphrase, and submit your email address. Follow then the procedure as you did when you created your account in the first place.
3.8. My email address changed. What can I do ?
Go to login.ceci-hpc.be/#change_email. Enter both the new and the previous email addresses. Follow then the procedure as you did when you created your account in the first place.
Note that emails ending with 'fundp.ac.be' have been translated as 'unamur.be' automatically.
3.9. Can I have more diskspace | more priority | sudo/root privileges ?
The general answer to that question is simply "sorry, no you can't" unless there has been an imponderable independent from your good will, e.g. a hardware crash.
However, do not hesitate to contact the support team and submit your problem. We might be able to find a solution that does not involve granting you privileges above those of the other users.
More specifically:
- diskspace is often limited in your home directory through quota. If you need more space than your quota allows, you should use scratch space.
- access to the cluster is shared among users. If you wish exclusive access to a whole cluster, be on the lookout for announcements for 'debugging sessions'.
- global software install is performed by the system administrators only, but you can always install software in your home directory.
3.10. Can I renew my account from abroad ?
Just like access to the clusters is restricted, access to the account renewal system is restricted to connections from inside the university networks. The solution is the same as when you want to connect to the clusters, namely either to use VPN, in which case you have nothing more to do than activate it, or an SSH gateway. In that latter case, you will simply run a command like
ssh -f -X my_login@thegateway.myuniversity.ac.be firefox -no-remote
to open a browser on the gateway and renew your account with that browser. You will probably need to copy/paste the links sent to you by email.
4. If this FAQ does not help...
4.1. I did not find an answer to my question here. What do I do ?
Have a look at the FAQ mentioned in the Clusters section of this website. Click on the link related to the cluster for which you have a specific question.
Also make sure to review the online documentation ; it might contain the answer you are looking for.
4.2. I still cannot find an answer.. Please help !
Then you can contact your local system administrator directly:
- UCL : The CISM team
- ULB : The ULB HPC centre
- ULiège : David Colignon
- UMons : Sébastien Kozlowskyj
- UNamur : Frédéric Wautelet
The best way to contact them is through the centralised support form
4.3 How can I make sure I get help quicky?
As a general rule, the more information you provide, the quicker you are likely to get a response. Please review the 5 tips to a better support request document for more information.