By Topic

Correctness within a constant of an optimal buffer allocation rule of thumb

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Anantharam, V. ; Dept. of Electr. Eng., Cornell Univ., Ithaca, NY, USA ; Ganesh, A.J.

The problem is to allocate a fixed number of buffers among the nodes of an open network of exponential servers with Bernoulli routing and Poisson arrivals so as to optimize some performance criterion associated with the time to buffer overflow, such as maximizing its mean or maximizing the probability that it exceeds some value. In earlier work, the authors used pathwise probabilistic arguments to derive a simple rule of thumb for this problem: allocate the buffers in inverse proportion to the logarithms of the effective service rates at the nodes. Effective service rate denotes the ratio of the service rate to the stationary arrival rate in the network with infinite buffers. They showed that this rule of thumb is accurate to within a known constant times the logarithm of the number of buffers as the number of buffers to be allocated becomes large. In the present paper, the authors use time reversal and Poisson clumping arguments to show that their rule of thumb is, in fact, much better than previously demonstrated. They show that the optimal buffer allocation is within a constant of the rule of thumb as the number of buffers to be allocated becomes large, although now they cannot estimate the constant. In numerical terms, the earlier result reduced the search space for the optimal buffer allocation from O(NJ-1) to O((log N)J-1), where J denotes the number of nodes and N the number of buffers to be allocated. The improvement reduces the search space to O(1)

Published in:

Information Theory, IEEE Transactions on  (Volume:40 ,  Issue: 3 )