Effect of Hotspot Traffic on Blocking Probability in Elastic Optical Networks
Paresh Upadhyay, Yatindra Nath Singh
TL;DR
This paper addresses how hotspot traffic influences blocking probability in Elastic Optical Networks (EONs). It models Routing and Spectrum Assignment (RSA) on the NSFNET topology using Poisson arrivals with rate $\lambda$ and exponential holding times with mean $1/\mu$, comparing uniform and hotspot traffic under contiguity and continuity constraints, with spectrum units of $W=12.5$ GHz and a guard band. By designating hotspot nodes based on Node Betweenness Centrality (NBC) and varying hotspot counts and traffic multipliers, it quantifies effects on the Request Blocking Probability ($RBP$) and Bandwidth Blocking Probability ($BBP$) using first-fit spectrum assignment (SP-FF) and hotspot variants (SP-FF-HP1, SP-FF-HP2). Key findings show that hotspot location, especially among high-NBC nodes, can significantly raise blocking, while increasing the number of hotspot nodes can mitigate blocking through more uniform load distribution; however, extreme hotspot traffic or closed hotspot regions can exacerbate blocking. The results offer practical insights for traffic engineering and spectrum management in EONs to improve reliability under imbalanced demand.
Abstract
In a circuit-switched network, traffic can be characterized by several factors that define how communication resources are allocated and utilized during a connection. The amount of traffic basically determines how frequently connection requests arrive, how long the setup connection remains active, and the bandwidth used. The Poisson Arrival Process models traffic arrival events at random intervals. It assumes that events happen independently of one another. This model is ideal for simulating traffic in networks where arrivals happen independently and randomly, such as the start of phone calls, data requests, or packet transmissions. The Poisson Arrival Process and uniformly choosing source and destination pair is been used most commonly by researchers to generate traffic in a network to test various promising routing and spectrum assignment algorithms. It checks the algorithm in uniformly loaded conditions and estimate its baseline performance. In real real-world scenario, a bunch of network nodes can start experiencing heavy data traffic compared to the rest of the network. This can lead to latency issues, or even outages if the network is not optimized to handle the load at these nodes which are also called hotspots. In other terms, hotspot in a network is an area or set of nodes within the network that have a higher likelihood of being involved in communication or data transmission compared to other areas. In this paper, we have tried to find what are the various factors involved in increasing the blocking probability in hotspot traffic scenarios. We have also compared the results with the uniform traffic load conditions in same topology.