Tandem Queue Architecture — Edge Resource Scaling

Modeled stochastic tandem queues for edge computing.

year: Jan 2026 — March 2026
role: Course Project
location: KTH Royal Institute of Technology
category: Academic

[diagram // tandem queues infrastructure]

Tandem queues infrastructure block diagram — two-stage tandem queuing network // M/M/1 wireless links → M/M/m edge cluster

Context

Evaluated end-to-end latency in a multi-user edge computing architecture modeled as a two-stage tandem queuing network (M/M/1 → M/M/m) to balance traffic load.

Methodology

Applied Burke's Theorem and the Erlang-C formula to mathematically model the joint impact of wireless link capacities and shared computational server resource scaling.

Analysis

Developed a MATLAB simulation framework to stress-test the system, analyzing total delays, stochastic waiting times, and stability regions across various server configurations (m = 1, 2, 4).

Outcome

Identified the critical performance threshold where system bottlenecks transition from wireless transmission boundaries (<20 Mbps) to an immutable computational processing floor (>60 Mbps).

[data // bottleneck analysis & latency floors]

Multi-server performance under shared load — fig.01 // multi-server performance under shared load (m = 1, 2, 4)

Impact of wireless transmission rate on total delay — fig.02 // impact of wireless transmission rate on total delay

tap any figure // enlarge

Queuing TheoryM/M/1 & M/M/m SystemsPerformance ModelingMATLAB

Real-Time Audio Equalizer

open →