Skip to main content
Optiscale
Log inContact us
Solutions
HPC Cluster Design · CPU supercomputersAI Supercomputer · GPU training & inference clustersParallel Storage · Spectrum Scale · Lustre · BeeGFS · VASTInterconnect · InfiniBand · RoCE fabricsSchedulers · Slurm · KubernetesApplication Workload · Reference diagrams by workloadSizing Calculator · 5-minute sizing + 5-year TCO
Products
Luxe Vision · Unified monitoring across heterogeneous resourcesLuxe Ray · Heterogeneous resource operations & unified controlLuxe Vantage · Usage & billing built on the AI/HPC schedulerLuxe Orbit · Heterogeneous software provisioning & parallel managementLuxe Series Overview · The unified 4-layer storyRequest a Live Demo · Vision · Ray · VantageRequest a Closed PoC · Orbit — 1:1 environment setup
Services
Architecture Consulting · Workload definition → specification designImplementation & PM · Vendor integration & project managementPerformance Tuning · Architecture & configuration optimization plus code-level performance gainsManaged Operations · Multi-year operations contracts
Resources
TCO Calculator · On-premises vs. the three major cloudsSelf-Check · 8-question workload assessmentApplication Workload Library · 6 workload diagramsWhite Papers · In-depth technical white papersBlog · Tech Notes · Engineering blogNewsletter · Biweekly infrastructure updates
BLOGProduct

Luxe Vantage Billing Design — How to Build Automatic Per-Department Chargeback

Can you get cloud-like billing on an on-premises GPU cluster? Slurm sacct + Kubernetes metrics + per-department rate policy.

Product Team··8 min read

Why This Product

Consulting on university GPU clusters, we heard the same question over and over:

> "Ten departments share one cluster — is there a way to settle costs without disputes?"

The conventional approach: at month's end, the ops team runs sacct and assembles the numbers in Excel. It takes days, and disputes break out between departments every month.

Luxe Vantage's Design Decisions

Decision 1 — Usage definition: not GPU-hours, but a weighted average of GPU, memory, and time.

If you treat a job holding a GPU for 24 hours at 10% utilization the same as a job holding it for 4 hours at 90% utilization — as a single GPU-hour each — the incentives break. We add `(allocated_gpu × wall_time) + α × (mem_peak / mem_total) × wall_time` on top of the base rate.

Decision 2 — Priority queues get a separate rate.

The urgent queue is 1.5× the standard queue. When a user requests priority, the system shows the added cost immediately. The result: only genuinely urgent jobs enter the urgent queue.

Decision 3 — Invoices are PDFs, but the data is always viewable.

Departments receive a month-end PDF, but they always see real-time running totals in Vision's finance team view. This prevents "surprise bill at month's end" incidents.

Implementation

- Data sources: Slurm sacct (batch) + Kubernetes (Volcano + Prometheus) + Spectrum Scale usage. - Rate policy: a matrix by department, project, and time-of-day. Importable via CSV. - Output: PDF + CSV + REST API (integration with finance systems).

Case Results

/case-studies/gpu-chargeback-multi-dept — 8 departments, 90% reduction in settlement work, department satisfaction 4.7/5.

Apply for the Beta

/products/luxe-vantage — closed beta in progress.

Luxe Vantage Billing Design — How to Build Automatic Per-Department Chargeback | Optiscale