Lab: Measure Inference Distribution — Benchmarking the Mesh

What you'll build

Vertical stack context: DOSFIUniversity (research) → DOSFI Academy (this course) → MeshNativeExchange (publishing & monetizing). Students progress from Learning → Building → Publishing → Monetizing. In this production-grade lab, you will perform a rigorous benchmarking study on the DOSFI distributed mesh. You will deploy an image classification job across varying node counts (1, 3, and 5) to observe latency, cost-efficiency in Distributed Intelligence Units (DIU), and systemic bottlenecks. This lab provides the framework for optimizing sovereign distributed deployments.

Skills you'll practice

• Implement a standardized inference job benchmark
• Measure performance delta across varying cluster sizes
• Calculate DIU consumption for heterogeneous workloads
• Analyze network and compute bottlenecks in a distributed mesh
• Construct a formal performance comparison matrix

Before you start

• Foundational knowledge of distributed systems
• Experience with command-line infrastructure tools
• Basic understanding of inference pipeline metrics

Lab steps5 steps

1. 1
   Defining the Benchmark Protocol
2. 2
   Execution and Analysis Framework
3. 3
   Scaling the Topology: 3-Node vs 5-Node Performance
4. 4
   Benchmarking Data Collection and Comparison Tables
5. 5
   Analyzing Bottlenecks via Performance Optimization Framework