Performance analysis of edge-PLCs enabled industrial Internet of things

Abstract

With the recent advancement in Industrial Internet of Things (IIoT), general programmable logic controllers (PLCs) have been playing more and more critical roles in industrial control systems (ICSs), such as providing local data processing, decentralized control and fault diagnosis. These so called edge-PLCs, directly receive the raw data from sensors embedded in factory equipments, put them into predefined memory space and perform analysis using programs such as the ladder logic. The challenge is how to allocate blocks in the fixed-size memory to different sensors so as to match irregular data flows. In this paper, we try to conduct performance analysis of different partition instances of the memory in the edge-PLC by modeling this problem as a multiple single-server queueing systems. We assume every sensing flow is independent of each other and has its dedicated processer. Changes can be made to partition instances to adapt to the external environment, such as the rising of order numbers or product category switching. Each state of the environment is defined by the finite state Markov chain and arrival of sensing data flows follow the stationary Poisson process. The data in the queue will expire after staying in the memory for a while. The duration of availability and service is modeled as the exponential distribution. The performance measured under different system states are analyzed in the simulation.