The Lab

Experiments, insights, and hands-on challenges for HPC engineers.

Deep technical explorations of SLURM scheduling, real cluster data analysis, and practical challenges that sharpen your HPC operations skills.

Featured

Deep Dive12 min read

Deep Dive: Understanding SLURM's Fair Share Algorithm

How SLURM calculates job priorities using historical usage patterns and why it matters for optimizing cluster throughput.

Fair ShareSchedulingPriority
Experiment8 min read

Experiment: GPU Utilization Patterns in Multi-Tenant Clusters

Real-world data from a 128-GPU cluster showing how job packing strategies affect overall utilization and user wait times.

GPUMulti-TenantOptimization
Deep Dive2025-01-28

Deep Dive: Understanding SLURM's Fair Share Algorithm

How SLURM calculates job priorities using historical usage patterns and why it matters for optimizing cluster throughput.

12 min read
Experiment2025-01-24

Experiment: GPU Utilization Patterns in Multi-Tenant Clusters

Real-world data from a 128-GPU cluster showing how job packing strategies affect overall utilization and user wait times.

8 min read
Challenge2025-01-20

Challenge: Debugging a Stuck Job Queue

Walk through identifying and resolving a complex queue blockage caused by misconfigured job dependencies.

6 min read
Tutorial2025-01-15

Tutorial: Setting Up Custom SLURM Accounting

Step-by-step guide to implementing fine-grained resource tracking for better cost allocation and cluster analytics.

15 min read
Deep Dive2025-01-10

Understanding SLURM Partition QOS Policies

How to leverage Quality of Service settings to create priority tiers and enforce resource limits across different user groups.

10 min read
Experiment2025-01-05

Benchmarking Job Schedulers: SLURM vs PBS vs LSF

Performance comparison of major HPC schedulers under various workload patterns, with real cluster data and insights.

14 min read

Ready to level up your SLURM skills?

Join early access to get notified when new experiments, challenges, and deep-dive articles are published.