Here is the current advice and instructions for applying for HECBioSim time on UK HPC resources.
You should be aware of the following information.
You must acknowledge EPSRC and HECBioSim in any outputs arising from any project awards granted, you can do this by using the following text:
This project made use of time on HPC granted via the UK High-End Computing Consortium for Biomolecular Simulation, HECBioSim (http://hecbiosim.ac.uk), supported by EPSRC (grant no. EP/R029407/1)
We are collecting information as to whether a given project aligns against a number of our current research themes. We would encourage applicants to indicate the correct theme/s for their project proposal, projects that align against these themes may be looked on as more desireable. Our current research themes are:
- Antimicrobial Resistance.
- Biomolecular design for synthetic biology, future health care and industry.
- Simulation for drug discovery.
- Understanding the molecular origins of disease.
- Pushing the boundaries of biomolecular simulation.
- Multiscale modelling of biomolecular systems.
- Directed and self-assembly
The following allocation period dates are the date ranges in which you will be able to get physical access to different HPC facilities via HECBioSim to make use of your resource allocation awarded. Note, that these dates are different to the panel deadline dates in which you should have made your application.
Current Allocation Period: 01/11/2020 - 30/04/2020
Next Allocation Period: Machine Closed
Current Allocation Period: TBA
Next Allocation Period: TBA
There are two main channels for getting time via HECBioSim. These are the smaller pilot type projects where you are wanting to try a new method, code or test a new architecture. Or there is a standard projects application that will go via peer review at a Resource Allocation Panel.
In order to apply for HECBioSim time on any of the UK HPC resources that we run calls for, please use the following process for the most efficient use of your time.
Step 1 ARCHER2 Only (if not applying to ARCHER2, please skip to step 2)
Determine whether you will need to complete an ARCHER2 technical assessment. Some programs are well known by the HEC consortium to scale well on ARCHER2, below is a list of programs that we know scale well and thus will not require a technical assessment when applying for HEC time through HECBioSim.
Complete the ARCHER2 technical assessment using the form here, this must be completed and handed to the ARCHER2 team to be signed off before proceeding with your application with us, otherwise your case will be rejected at the panel.
Please note: It is possible if you are using new code that is already tested via a differnt route that you can apply for HECtime through HECBioSim using a previously accepted technical assessment.
Step 2 - Justification Document
You should prepare a document, no more than 2 pages. To justify your scientific case along with technical information about your proposed project, this is what will be assessed at the RAP. We have included some advice for this here, we strongly recommend that you read this!
We recommend that you use our benchmark results to help you justify your time allocations and choice of architecture!
Step 3 - Apply
You should now have everything that you will need to apply for time on any of the machines via HECBioSim. Fill out the relevant application form here and submit your application.
Your project will be submitted to the next consortium panel for review. Wait for a response from the consortium panel which will detail whether or not you have been successful and will include further instructions on setting up your account. Please note that the decision may take up to one month after the date of the corresponding deadline, this is to allow the panel to review all applications.
This advice is aimed to help improve the quality of applications and help to level the playing field between more established groups and those that are new to our community. These guidelines aim to show some of the characteristics that all strong applications have in common whilst addressing some of the common questions we are asked. The justification document should be made up of two main parts, with the science case forming the bulk of it and the technical case to show a breakdown of how the resources are to be used.
You should explain the science behind the study you are apply for HPC time for. This should include a brief overview of the aims of the project, what you will be studying and what the expected impact of this study would be, we would encourage you to use imagery and to summarise any preliminary work to help justfy your case. If you are working in very sensitive areas, and cannot for various reasons include some details in your application, this does not mean you are excluded from applying but we would ask you to get in touch so that we can advise on how best to approach this document.
We would recommend that you detail the simulations you are doing. This should include information on the software and methods you are aiming to use in your simulations. If you are doing quite a few systems or multiple codes then it is best to tabulate this information giving details on how many nanoseconds each run, how many repeats for each system for instance.
When choosing the machine architecture for your simulations, you should justify your choice of resource (ie why Tier1 and not Tier2). For example, if you are using AMBER, then this code runs overwhelmingly strong upon GPU based architectures, so if you were to apply for ARCHER2 which has no GPUs and poor scaling for AMBER then we would be looking for instance for if you are using methods within AMBER that do not have GPU support yet. Another example is if you are doing a type of study that is massively parallel, and thus requires access to a high speed interconnect.
You can increase your chances of obtaining at least some allocation by providing a reduction scenario. These panels are often oversubscribed, sometimes by nearly 2X the amount of resource we can allocate. Our panels are usually in a position where they are considering reduction scenarios, which can include allocation on a different resource or to cut the time given or to reject the case entirely. It often makes this process better for the applicant and the panel if reductions can be done in a way that is not detrimental to the study, ie that at least some progress can be made and potentially realise some outputs. Itemising as suggested above is one natural way to do this as you can then simply suggest a priority or a minimum as to what could be done under a reduction scenario. Another is if you could reduce the lengths or numbers of replicas etc.
We will add answers to questions that are most commonly sent to us here.
In short, no. The PI must be a permanent member of academic staff affiliated to a UK institution. Proposals sent by PhD students or post doctoral researchers will be rejected.
No, users can be of any nationality but must be working/studying at a UK institution.
We are unable to offer time directly to industrial projects, nor can we sell access or act as a commercial service provider. We do however welcome applications from collaborative projects between academic and industrial parties, often such projects will be ranked highly due to having an industrial collaborator. These collaborations are highly encouraged, do not require financial contribution from the industrial partner and can be quite broad in scope (methods development, test applications, studentships, code development, to try out ideas, etc).
A project via HECBioSim can last no longer than 6 months.
The ARCHER2 technical assessment process is designed to ensure that codes are used effectively on this resource (scaling, job sizes and durations, etc.). In the case of the common MD codes, being used in “standard” ways, the consortium has sufficient existing knowledge of their performance to be able to by-pass this step. The current list of software that does not require you to get an ARCHER2 technical assessment is:
Should you feel that you know of software that should appear on that list then please get in touch and explain why you think it should be on the list. It may arise that the consortium could utilise your knowledge/expertise with such software such that we can broaden our support for software.
New project applications from applicants that have had previously successful applications and then subsequently went on to under use their allocation by 50% and upwards at the time the project closed, will be looked at. Project applications that fall into this category will likely be heavily penalised during the panel review.
Unfortunately due to the limited amount of compute time at each HECBioSim panel and the high volume of applications for HECBioSim time on HPC, it is not always possible to accept every single application. If your application is rejected then your options include but are not limited to:
- Submit another application to HECBioSim for the next decision panel (it will be subject to the same review process).
- You can apply for various types of access to ARCHER2 directly from the ARCHER2 website.
- If your institution is a partner on one of the Tier2 facilities then you could apply through your local contact if they have capacity to offer this.
We typically do this on the machines that are shared by many communities, user groups and research councils. This is because what typically happens is that we see a lot of congestion towards the end of an allocation period and jobs can't be run due to all users groups having to use their resource at the same time. To prevent this happening, we run active profiling of the resource use, and we require you to burn the first half of your time both before the half way point in the allocation period and in order to release the second half. For the worst offenders, the second half may be reprofiled if use is poor.
On ARCHER2 the units you will see in your safe account are the CU (Compute Unit), this compute unit simply refers to the use of 1 compute node for 1 hour of compute. So you simply multiply the number of nodes and the number of hours of use to get the total CU for a simulation. On the Tier2 machines such as JADE/JADE2 and Bede, you will the units kAU in SAFE, these units are set to mean 1 GPU.hr, so you would multiply the number of GPUs used by the simulation by the number of hours to get the total kAU for a given simulation. The GPU.hr is simply the number of GPUs used for a given number of hours, so for example on JADE2 if you used 4 GPUs (there are 8 on a node) on a node for 3 hours then your simulation used 12 GPU.hrs.
We will always have the application form set to close at 17:00 (GMT) on the day of the application deadline, this is inline with UK business hours. The form link will disappear precisely at 17:00 and applications already underway will not be able to be submitted past 17:00, it is strongly advised to complete your application in the weeks that is open.