Technical Requirements¶
To deploy an application on Game Warden, your solution must satisfy the following architecture and security specifications before you engage the Game Warden onboarding team.
Architecture¶
| Requirement | Details |
|---|---|
| Containerization (OCI-compliant) | The application must run in containers that conform to the Open Container Initiative (OCI) specification. |
| Kubernetes compatibility | The application must be deployable on Kubernetes, using standard Kubernetes primitives (Deployments, StatefulSets, Services, ConfigMaps, etc.) and must not rely on host-level access or non-Kubernetes runtimes. |
| Database seeding | Provide automated seed services or SQL/DDL scripts for the Game Warden team to execute. At IL4 you will not have direct write access to production databases. |
| CPU architecture | Workloads must target AMD64/x86_64 or ARM64/AArch64. |
Security¶
| Requirement | Details |
|---|---|
| Meeting ATO vulnerability baseline | Game Warden performs continuous security scans. All findings must be remediated in accordance with the Acceptance Baseline Criteria. Components must be patched regularly to maintain ATO compliance. |
| Continuous CVE remediation | New CVEs discovered post-deployment must be resolved promptly by the application team. |
| DoW-approved authentication | Applications must integrate with a DoW-approved identity provider—Game Warden SSO or Platform One SSO. |
| Credentialed access (IL4+) | Personnel accessing IL4+ environments require a valid government access card credential obtained through standard DoW vetting. |
| Data classification limits | Permitted data classifications: CUI, PII, IL2, IL4, IL5, ITAR. Contact Game Warden before processing IL6, Special Access Programs (SAP), or Sensitive Compartmented Information (SCI) data. |
AWS GPU support by environment¶
For a list of Amazon EC2 instance types available in AWS GovCloud (US-East), see the AWS documentation.
| EC2 Instance | Instance Name | GPU Supported |
|---|---|---|
| g3 | g3.4xlarge | 1 NVIDIA Tesla M60 GPU, with 2048 parallel processing cores and 8 GiB of video memory |
| g3.8xlarge | 2 NVIDIA Tesla M60 GPUs, each with 2048 parallel processing cores and 8 GiB of video memory | |
| g3.16xlarge | 4 NVIDIA Tesla M60 GPUs, each with 2048 parallel processing cores and 8 GiB of video memory | |
| g4dn | g4dn.xlarge | 1 NVIDIA T4 Tensor Core GPU |
| g4dn.2xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.4xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.8xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.12xlarge | 4 NVIDIA T4 Tensor Core GPUs | |
| g4dn.16xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.metal | 8 NVIDIA T4 Tensor Core GPUs | |
| g5 | g5.xlarge | 1 NVIDIA A10G Tensor Core GPU |
| g5.2xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.4xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.8xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.16xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.12xlarge | 4 NVIDIA A10G Tensor Core GPUs | |
| g5.24xlarge | 4 NVIDIA A10G Tensor Core GPUs | |
| g5.48xlarge | 8 NVIDIA A10G Tensor Core GPUs | |
| p3 | p3.2xlarge | 1 NVIDIA Tesla V100 GPU, pairing 5,120 CUDA Cores and 640 Tensor Cores |
| p3.8xlarge | 4 NVIDIA Tesla V100 GPUs, each pairing 5,120 CUDA Cores and 640 Tensor Cores | |
| p3.16xlarge | 8 NVIDIA Tesla V100 GPUs, each pairing 5,120 CUDA Cores and 640 Tensor Cores | |
| p3dn | p3dn.24xlarge | 8 NVIDIA Tesla V100 GPUs |
| p5 | p5.48xlarge | 8 NVIDIA H100 Tensor Core GPUs |
| p5en.48xlarge | 8 NVIDIA H100 Tensor Core GPUs |
| EC2 Instance | Instance Name | GPU Supported |
|---|---|---|
| g3 | g3.4xlarge | 1 NVIDIA Tesla M60 GPU, with 2048 parallel processing cores and 8 GiB of video memory |
| g3.8xlarge | 2 NVIDIA Tesla M60 GPUs, each with 2048 parallel processing cores and 8 GiB of video memory | |
| g3.16xlarge | 4 NVIDIA Tesla M60 GPUs, each with 2048 parallel processing cores and 8 GiB of video memory | |
| g4dn | g4dn.xlarge | 1 NVIDIA T4 Tensor Core GPU |
| g4dn.2xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.4xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.8xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.12xlarge | 4 NVIDIA T4 Tensor Core GPUs | |
| g4dn.16xlarge | 1 NVIDIA T4 Tensor Core GPU | |
| g4dn.metal | 8 NVIDIA T4 Tensor Core GPUs | |
| g5 | g5.xlarge | 1 NVIDIA A10G Tensor Core GPU |
| g5.2xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.4xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.8xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.16xlarge | 1 NVIDIA A10G Tensor Core GPU | |
| g5.12xlarge | 4 NVIDIA A10G Tensor Core GPUs | |
| g5.24xlarge | 4 NVIDIA A10G Tensor Core GPUs | |
| g5.48xlarge | 8 NVIDIA A10G Tensor Core GPUs | |
| g6 | g6.xlarge | 1 NVIDIA L4 Tensor Core GPU |
| g6.2xlarge | 1 NVIDIA L4 Tensor Core GPU | |
| g6.4xlarge | 1 NVIDIA L4 Tensor Core GPU | |
| g6.8xlarge | 1 NVIDIA L4 Tensor Core GPU | |
| g6.16xlarge | 1 NVIDIA L4 Tensor Core GPU | |
| g6.12xlarge | 4 NVIDIA L4 Tensor Core GPUs | |
| g6.24xlarge | 4 NVIDIA L4 Tensor Core GPUs | |
| g6.48xlarge | 8 NVIDIA L4 Tensor Core GPUs | |
| p3 | p3.2xlarge | 1 NVIDIA Tesla V100 GPU, pairing 5,120 CUDA Cores and 640 Tensor Cores |
| p3.8xlarge | 4 NVIDIA Tesla V100 GPUs, each pairing 5,120 CUDA Cores and 640 Tensor Cores | |
| p3.16xlarge | 8 NVIDIA Tesla V100 GPUs, each pairing 5,120 CUDA Cores and 640 Tensor Cores | |
| p3dn | p3dn.24xlarge | 8 NVIDIA Tesla V100 GPUs |
| p4d | p4d.24xlarge | 8 NVIDIA A100 Tensor Core GPUs |
| p5 | p5.48xlarge | 8 NVIDIA H100 Tensor Core GPUs |
GCP GPU support¶
Below are the GPUs supported in the us-east4 region (Northern Virginia) for Assured Workloads:
| GPU Model | Machine Series | Typical Use Case |
|---|---|---|
| NVIDIA H100 (80GB) | A3 | Large-scale AI training and LLM fine-tuning. |
| NVIDIA A100 (40GB/80GB) | A2 | High-performance deep learning and data science. |
| NVIDIA L4 | G2 | AI inference, video processing, and smaller training tasks. |
| NVIDIA T4 | N1 | Cost-effective inference and graphics acceleration. |
| NVIDIA RTX PRO 6000* | G4 | High-end workstations and Blackwell architecture tasks. |
Azure GPU support¶
Below are the GPUs supported in Azure Government Cloud (US Gov Virginia):
| Series | GPU Model | Primary Use Case |
|---|---|---|
| ND H100 v5 | NVIDIA H100 | High-End AI: LLM training and massive Generative AI. |
| ND A100 v4 | NVIDIA A100 (80GB) | Deep Learning: Large-scale training and high-memory HPC. |
| NCads H100 v5 | NVIDIA H100 | Inference/Training: Focused on PCIe H100 performance. |
| NCas T4 v3 | NVIDIA T4 | Inference: Lightweight AI, video encoding, and data processing. |
| NVads A10 v5 | NVIDIA A10 | Visualization: Graphics-heavy apps, CAD, and VDI. |
| NCv3 / NCv2 | NVIDIA V100 / P100 | Legacy Compute: Older HPC workloads (often being phased out). |
Programming language¶
Game Warden does not impose restrictions on your choice of programming language or framework. You can deploy your application in any language, as long as it is packaged as a Linux-based container. Note that Windows-based containers are not supported.
Next steps¶
Confirm your architecture and security posture meet the requirements above, then contact the Game Warden team. We’ll help ensure alignment and support your application launch. Reach out to the Growth team for details.