Project Proposal (PP)



Resource Computing Allocation Committee

RCAC Cloud Proposal (RCP)

|Project Title | |

|Principal Investigator (PI) | |

|PI e-mail | |

|PI phone | |

|PI organizational affiliation | |

|PI signature | |

| | |

| |By submitting this proposal, I approve the entire content of this document and |

| |associated supporting documents. |

|Date of Proposal | |

|CRG, CCF, KACST, industrial and/or | |

|governmental funding (award details with funds| |

|and duration) related to the project | |

|Amount requested in USD | |

|Have you used commercial Cloud | |

|resources in the past? If so, please | |

|state the total amount spent from | |

|KAUST resources. | |

Available System:

This is a request form for University sponsorship of commercial cloud use.

Limitation of Validity:

The charging authority granted by a successful application must be used only for the project described below.

Submission:

Please e-mail the completed proposal to the RCAC recording secretary bilel.hadri@kaust.edu.sa

Additional KAUST Investigators/Users (add more blocks as needed)

|1 |Name: | |

| |Email: | |

| |Tel: | |

| |Organisation: | |

|2 |Name: | |

| |Email: | |

| |Tel: | |

| |Organisation: | |

Proposed External Collaborators (add more blocks as needed):

|1 |Name: | |

| |Email: | |

| |Tel: | |

| |Organisation: | |

| |Address: | |

Project Description:

Please provide a short description of the project, including current state of art, research work proposed, expected milestones, and deliverables, including target publication venue(s). Please mention potential impact, including alignments with KAUST strategic initiatives or translational opportunities. Examples of criteria for why a campaign cannot be run on Ibex and must incur operational spending on the cloud might include these:

1) need other types of hardware not present in Ibex (processor, memory configuration, encryption configuration, etc.) or software that cannot easily be installed on Ibex

2) need more nodes than can be allocated simultaneously on Ibex within a reasonable queueing time

3) need cloud or distributed resources as part of the research proof of concept (e.g., federated learning)

4) simply not enough capacity, as evidenced by extrapolated Ibex usage data combined with CRR/SRR approved requirements

5) requires too much data to be moved and stored locally

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

| |

List of Publications:

Please provide a list of publications that have appeared archivally or been accepted for journals, conference proceedings, posters, etc., resulting from GPU resources from any previous allocations on Ibex or using Cloud resources since 2020 (inclusive).

| |

Training Purposes:

Please briefly describe any relevant training that will be enabled by this proposal, including the names of KAUST Doctoral or Master’s students, if applicable

| |

| |

| |

| |

Typical Execution Description:

Please briefly describe the job characteristics of the runs, such as the type of processing required, the number of GPUs per task, expected run time of the jobs, and expected total GPU hours required.

| |

| |

| |

| |

Total Resource Requirement Justification:

Please detail how the USD request was calculated.

Example: “We expect to consume 10K GPU hours in the campaign described above. The average cost per GPU hour on Amazon is US$4.2 on A100 GPUs. Therefore, the projected cost is around 10K * 4.2 = US$ 42K. Considering additional costs such as data storage and transfer, the total request is US$45K.”

| |

| |

| |

| |

| |

| |

| |

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download