====== Resource Allocations and Project Requests ====== This page explains how HPC resources are allocated, how usage is measured, and how to request additional resources for research projects. Most users can start working immediately using the default allocation. Larger projects, GPU-intensive workloads, and research collaborations may require additional resources. ===== Resource Allocations ===== Resource allocations define how much of the HPC infrastructure a user or project may consume over time. Resources are measured as: * CPU-hours * GPU-hours * Storage * Project allocations A CPU-hour means one CPU core used for one hour. Examples: ^ Resources Used ^ Consumption ^ | 1 CPU for 1 hour | 1 CPU-hour | | 4 CPUs for 1 hour | 4 CPU-hours | | 8 CPUs for 2 hours | 16 CPU-hours | | 1 GPU for 1 hour | 1 GPU-hour | | 4 GPUs for 2 hours | 8 GPU-hours | ===== Default Allocations ===== New users receive a default allocation based on their affiliation. ^ User Type ^ CPU Hours / Month ^ GPU Hours / Month ^ | VU MIF Users | 2000 | 100 | | Other Vilnius University Users | 100 | 10 | | External Users | Project-specific | Project-specific | The default allocation is intended for coursework, learning, testing, and smaller research workloads. ===== How Resource Usage Is Calculated ===== CPU and GPU usage are accumulated for all completed jobs. Example: 4 CPU cores × 3 hours = 12 CPU-hours Example: 2 GPUs × 5 hours = 10 GPU-hours Resource usage is based on allocated resources, not the actual resource utilization of an application. For example, if a job reserves 8 CPU cores but actively uses only 4, accounting will still charge all 8 allocated CPU cores. ===== Viewing Resource Usage ===== If your jobs stop starting because of: AssocGrpCPUMinutesLimit or AssocGrpGRESMinutes you may have reached your allocation limit. ===== Method 1: View Usage with sreport ===== Display CPU and GPU usage for the current month: sreport -T cpu,mem,gres/gpu cluster AccountUtilizationByUser \ Start=$(date +%m01) End=$(date +%m31) User=$USER Resource usage is reported in minutes. Divide by 60 to obtain hours. ===== Method 2: View Allocation Usage with sshare ===== Display allocation information: sshare -l -A USERNAME_mif -p \ -o GrpTRESRaw,GrpTRESMins,TRESRunMins Replace ''USERNAME'' with your HPC username. Fields: ^ Field ^ Description ^ | ''GrpTRESRaw'' | Resources already consumed | | ''GrpTRESMins'' | Resource limit | | ''TRESRunMins'' | Resources currently consumed by running jobs | ===== When Should You Request More Resources? ===== Consider requesting additional resources when: * Your monthly CPU allocation is insufficient * Your monthly GPU allocation is insufficient * Your project needs dedicated resources * You collaborate with multiple researchers * You need long-running simulations * You run machine learning or AI workloads * You need large data storage allocations Typical examples: * Scientific simulations * Machine learning model training * Large-scale data analysis * Research collaborations * Grant-funded projects ===== Request Additional Resources ===== Additional resources can be requested through the IT Research Center service portal: [[https://mif.vu.lt/lt3/en/about/structure/it-research-center#ordering|IT Research Center (ITRC) Service Requests]] When submitting a request, it is helpful to provide: * Project title * Principal investigator or supervisor * Project duration * Estimated CPU usage * Estimated GPU usage * Storage requirements * Number of project members Providing accurate information helps administrators assign resources more efficiently. ===== How Allocation Increases Work ===== Resource increases are normally granted for the current month and future months. Example: * Default monthly limit: 100 GPU-hours * Already consumed this month: 40 GPU-hours * New approved limit: 600 GPU-hours Additional allocation granted: 600 - 40 = 560 GPU-hours available The new monthly limit in future months becomes: 600 GPU-hours/month Unused resources do not carry over to the next month. ===== Project Allocations ===== Larger allocations are usually managed through projects. Project allocations make it possible to: * Share resources between team members * Track project resource consumption * Assign dedicated CPU and GPU quotas * Manage collaboration across institutions Jobs submitted under a project allocation use: #SBATCH --account=alloc_xxxx_project Example: #!/bin/bash #SBATCH --account=alloc_xxxx_project #SBATCH --partition=main #SBATCH --cpus-per-task=4 #SBATCH --time=01:00:00 python3 script.py ===== Waldur Projects ===== Project allocations are managed through Waldur. Waldur provides: * Project management * Resource allocations * Usage accounting * User invitations * Collaboration management For details see: * [[en:hpc-new:waldur|Waldur User Guide]] ===== Resource Efficiency ===== Efficient resource usage helps everyone access the cluster more quickly. Recommended practices: * Request only the CPU cores you need * Request only the memory you need * Set realistic runtime limits * Cancel jobs that are no longer needed * Use GPUs only when required Avoid: * Reserving large numbers of unused CPU cores * Reserving GPUs that remain idle * Excessively long walltime limits * Leaving large jobs running unnecessarily A smaller and more accurate resource request often starts faster than an oversized job. ===== Frequently Asked Questions ===== ==== Why won't my jobs start? ==== Check whether you have exceeded your allocation limits. Common errors include: AssocGrpCPUMinutesLimit AssocGrpGRESMinutes These usually indicate that additional resources are required. ==== Do VU MIF users need a project allocation? ==== Usually not. Most VU MIF users can work using their default monthly allocation. ==== Do I need the --account option? ==== Most VU MIF users do not. Only users working within a project allocation should use: #SBATCH --account=alloc_xxxx_project ==== Can allocations be shared? ==== Yes. Project allocations can be shared between project members. ==== Do unused resources carry over to the next month? ==== No. Allocations are renewed according to the approved monthly limits. ===== Common Problems ===== ==== AssocGrpCPUMinutesLimit ==== Your CPU allocation has been exhausted. Check current usage and request additional resources if necessary. ==== AssocGrpGRESMinutes ==== Your GPU allocation has been exhausted. Check current usage and request additional resources if necessary. ==== Invalid account specified ==== Verify that: * The allocation exists * You have been added to the project * The allocation name is spelled correctly ===== Next Steps ===== Now that you understand how resources are allocated and managed, continue with: * [[en:hpc-new:waldur|Waldur User Guide]] You may also revisit: * [[en:hpc-new:getting_started|First Successful Job]] * [[en:hpc-new:slurm|Batch Job Execution (SLURM)]] ---- This guide was prepared and updated by the VU MIF HPC team with the assistance of Microsoft Copilot. All procedures and examples should be validated against the current HPC environment, software stack, and operational policies.