DeathStarBench is a realistic, end-to-end microservices benchmark suite that models interconnected services (e.g., social network, e-commerce) to stress modern cloud native stacks. It reproduces real transaction flows to expose latency, throughput, and interaction patterns across services.

Core Components:

• Representative microservice workloads (user-facing services, databases, and caches).
• Workload drivers that simulate realistic traffic patterns and end-to-end transactions.
• Integrated metrics, tracing, and tail-end behavior.
• Deployment for VMs, containers, and Kubernetes clusters Why is Death Star Bench important?

DeathStarBench provides repeatable, realistic scenarios to measure how architecture, orchestration, and hardware impact application-level performance – especially tail latency and cross-service behaviors. Key advantages:

• Reveals systemic performance issues that unit or single-service tests miss
• Facilitates apples-to-apples comparisons of hardware, networking, and orchestration choices.
• Guides tuning, capacity planning, and cost-performances trade-offs for real microservice workloads.

• Ampere Performance Blog
• Ampere Case Studies
• Markov Chain Decomposition Based On Total Expectation Theorem