AWS Architecture & Cost Benchmark
Budget: €8 – €30 EUR
I need a seasoned AWS architect who can, within the next 24 hours, help me decide whether to keep our CPU- and GPU-intensive workloads at Hetzner or migrate to AWS for better performance. The immediate objective is a clear, data-backed recommendation focused on performance improvement; if the numbers prove compelling, we will move straight into Phase 2 (implementation) tomorrow.
Here is what I expect from Phase 1:
• A side-by-side cost and performance comparison of Hetzner versus AWS, covering on-demand, Savings Plan and Spot pricing for EC2, projected container costs on ECS/EKS, and any Lambda-based serverless bursts we might leverage.
• A proposed high-level architecture that shows which workloads belong on EC2, where container orchestration under EKS/ECS makes sense, and the limited cases where Lambda would cut latency or cost.
• Concrete guidance on instance families for both CPU-heavy analytics and GPU-accelerated model inference, including expected vCPU/GPU utilisation numbers.
• A brief migration timeline and risk outline so I can green-light Phase 2 instantly.
Acceptance criteria: the comparison must use current public pricing, specify regional assumptions, highlight any hidden network or storage premiums, and be delivered as a concise deck plus a spreadsheet I can share with the finance team.
If your analysis convinces us, be prepared to start building the environment right away.
Here is what I expect from Phase 1:
• A side-by-side cost and performance comparison of Hetzner versus AWS, covering on-demand, Savings Plan and Spot pricing for EC2, projected container costs on ECS/EKS, and any Lambda-based serverless bursts we might leverage.
• A proposed high-level architecture that shows which workloads belong on EC2, where container orchestration under EKS/ECS makes sense, and the limited cases where Lambda would cut latency or cost.
• Concrete guidance on instance families for both CPU-heavy analytics and GPU-accelerated model inference, including expected vCPU/GPU utilisation numbers.
• A brief migration timeline and risk outline so I can green-light Phase 2 instantly.
Acceptance criteria: the comparison must use current public pricing, specify regional assumptions, highlight any hidden network or storage premiums, and be delivered as a concise deck plus a spreadsheet I can share with the finance team.
If your analysis convinces us, be prepared to start building the environment right away.