Answers are grounded in this paper and cite their sources.
Graph representation aids the evaluation of microservice dependencies, yet a distinction is required between structural links and performance links that are enforced by simulator. A documented simulation experiment was conducted with 1,350 records from 450 synthetic architecture instances in three undirected topology families, five sizes (20–100 services), 30 seeds and three arrival rates. When the two parameters are computed, they satisfy the identity with no arbitrary constant. Average path length latency is related with r = 0.9994 and adjusted coefficient of 2.045 ms per SD (95% CI: 2.035 to 2.055). Confirming the additive model is that the latency follows 5(hops + 1). Based on a controlled experiment that used approximate edge matching, Scale-Free instances have the lowest latency for every size. The edge-adjusted advantage of Scale-Free instances over Clustered was between −1.70 and −2.97 ms (Holm p < 1e-10). The offered rate is tracked by the throughput; no saturation is established.
No takes yet. Share an insight, caveat, or question.
Abbas Ghaleb Talib (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: