China
Africa
Explore chapters and articles related to this topic
ATM Signaling and Traffic Control
Published in P. S. Neelakanta, ATM Telecommunications, 2018
Among the algorithms considered above, the equivalent capacity method achieves the highest level of statistical multiplexing while guaranteeing the desired CLR application at the link. The heavy traffic approximation becomes effective as the ratio of the buffer size to the burst length increases. If the buffer size is small, then it coincides with the deterministic allocation namely, the peak bandwidth allocation. The nonparametric approach is, in general, less sensitive to buffer size. In particular, increasing the buffer size may not result in a significant increase in the amount of statistical multiplexing that can be achieved with the nonparametric approach.
QoS-aware energy-efficient workload routing and server speed control policy in data centers: A robust queueing theoretic approach
Published in IISE Transactions, 2023
Seung Min Baik, Young Myoung Ko
The fundamental difference between the heavy-traffic approximation and robust queueing theory lies in the treatment of random variables. The heavy-traffic approximation first derives the distribution of the target performance measure, such as response time. Then, it calculates the quantile of the random variable using a cumulative distribution function. The robust queueing theory reverses this order. We compare the robust queueing theoretic approach to the heavy-traffic approximation used by Ko and Cho (2014) that approximates the response time process as a reflected Brownian motion using the steady-state results for the queue in Zhang and Zwart (2008): where is a virtual response time and is a reflected Brownian motion with drift and variance We refer the readers to Section 3 in Ko and Cho (2014) for a detailed explanation of other parameters. The approximate probability of exceeding a threshold, δ, is given by: The QoS constant ε is treated at the last moment.