Performance analysis of finite-buffered multistage interconnection networks

Sticky States in Banyan Network Queues and Their Application to Analysis

“…This is true over a variety of queue sizes, network sizes, and offered traffic rates. The throughput and latency for (8,2,4) networks with various arrival rates and queue sizes obtained through simulation and with Yoon's and sticky analyses appears in Table 3. The throughput obtained with the sticky model is closer to simulated values than Yoon's model for all simulations performed.…”

“…Theimer, et al [16] analyzes only single-buffered networks. The results presented in [4] lacked detail, probably because of space restrictions. As will be illustrated below, the queue state distributions obtained using Mun's analysis do not match those obtained from simulation as well as the analysis presented here, especially under light traffic, and so may be less useful in understanding congestion.…”

“…The rationale is that a head packet having been blocked will more likely be blocked again in the next cycle. Such approaches have been tried by Lin and Kleinrock [11], Mun and Youn [12], and Hsiao, et al [4] for finite queues and earlier by Theimer, et al [16] for the single-buffered networks.…”

“…The analyses presented by Mun and Youn [12], Theimer, et al [16], and Hsiao, et al [4] model each queue with several sets of states. Each set corresponds to the fate of the head packet in the previous cycle.…”

Congested Banyan network analysis using congested-queue states and neighboring-queue effects

“…The rationale is that a blocked head packet will more likely be blocked again in the next cycle. Such approaches have been used by Lin and Kleinrock [11], Mun and Youn [12], and Hsiao et al [4] for finite queues and earlier by Theimer et al [14] and Hsiao et al [5] for single-slot queues (single-buffered networks).…”