Access

Project application guide

What to prepare before you submit a Medium or Large project application for Marlowe. Note the job-profile, storage, and PDF requirements below — the application form has more detailed instructions.

Section 1

Project & PI information

Prepare basic project details: project title, abstract, PI name, group-member SUNet IDs, and any urgent timelines (for example, an upcoming grant or conference deadline).

Section 2

Project PDFs

Computational Suitability Statement (CSS)

Upload a 2–4 page PDF describing your computational needs, readiness, scalability, and justification for using Marlowe. For Medium projects, focus on computational suitability rather than a full scientific narrative. Please include:

  • A brief scientific overview for context
  • Prior experience on Marlowe or similar GPU-based systems
  • Any weak or strong scaling studies, completed or planned
  • Codes and toolchains used — include GitHub links where available, with instructions on how to run your codes
  • Job profiles: wall time, GPUs/node, concurrency, memory, I/O, and checkpointing
  • Computational readiness and tuning status, expected or demonstrated MFU, and similar metrics

Provide detail on why your workload requires more than standard lab or cloud resources. Large projects (more than 10,000 GPU-hours) should provide strong evidence of readiness and scalability.

Section 3

Computational profile

Summarize the typical and maximum job type, wall time, GPU and node usage, concurrency, and usage pattern for the project. This helps us assess fit with system capabilities.

Section 4

Technical requirements & feasibility

List the software frameworks, container tools, and any special configuration needs.

Section 5

Storage & data

Estimate your scratch storage capacity requirements, and indicate how data will be sourced, moved, and stored. Include details on checkpointing, if applicable.

Section 6

Impact & acknowledgments

Briefly describe expected outcomes, such as publications or software.

Review process

How applications are reviewed

Staff review the Computational Suitability Statement for both Medium and Large projects, considering:

  • Experience — familiarity with GPU clusters, including past experience
  • Software & resource requirements — codes, toolchains, clear instructions for running applications, and job profiles for GPU, CPU, memory, I/O, and checkpointing use
  • Computational readiness — scaling, optimization, and computational efficiency (for example, MFU)
  • Need for Marlowe — justification for needing GPU resources beyond a standard lab, cloud environment, or Sherlock
Start a project application →

Requires Basic Access first. See access tiers.