System Requirements

QScan is available in two variants: CPU-only and GPU-accelerated. Choose based on your throughput requirements and infrastructure.

CPU Variant

Resource	Minimum	Recommended
vCPUs	2	6
Memory	12 GB	24 GB
Disk	10 GB	20 GB

The ML models consume approximately 8.4 GB of memory at runtime. The recommended configuration supports running 2 pollers and 2 scanners concurrently.

GPU Variant

Resource	Minimum	Recommended
vCPUs	4	4
Memory	16 GB	16 GB
GPU	1x NVIDIA (CUDA-compatible)	1x NVIDIA L4 or better
Disk	10 GB	20 GB

GPU acceleration significantly improves inference throughput. Any CUDA-compatible NVIDIA GPU with sufficient VRAM is supported.

Network Requirements

Outbound connectivity (required):

• Pulse API: api-pulse.qpoint.io (TCP 443/HTTPS)

• S3 storage: Your configured S3 endpoint (e.g., s3.amazonaws.com, a MinIO instance, or s3.warehouse.qpoint.io for Qpoint-managed storage)

No inbound connections required. QScan only makes outbound connections to poll for jobs and retrieve artifacts. No ports need to be opened for inbound traffic.

Container Runtime

QScan is distributed as a container image and requires one of the following:

• Docker Engine 20.10+

• Kubernetes 1.24+

• Google Cloud Run

• Any OCI-compatible container runtime

Image Registry

The QScan container image is hosted at:

us-docker.pkg.dev/qpoint-edge/public/qscan