Download presentation
Presentation is loading. Please wait.
Published byGlenna Halim Modified over 6 years ago
1
The Variance of Production Counts over a Long Time Horizon
Yoni Nazarathy* EURANDOM, TU/e Contains joint work with Ahmad Al-Hanbali, Yoav Kerner, Michel Mandjes, Gideon Weiss and Ward Whitt Workshop on Stochastic Models of Manufacturing Systems Eindhoven, June 2010 *Supported by NWO-VIDI Grant of Erjen Lefeber
2
Problem Domain: Queueing Output Processes
PLANT OUTPUT - Single Server Queues - Tandem Queues - Re-Entrant Lines Desired over long term: High Throughput Low Variability Our focus: for large T
3
Asymptotic Variance Rate of Outputs For Renewal Processes:
Variance Curves Example: Stationary stable M/M/1, D(t) is PoissonProcess( ): Example: Stationary M/M/1/1 with D(t) is RenewalProcess(Erlang(2, )): Asymptotic Variance Rate of Outputs For Renewal Processes:
4
Asymptotic Variance Rate
M/M/1 Non-Stop Service Burkes Theorem
5
The Basic Loss-Less Stable Queueing System
Q(t)
6
Our main focus: Overloaded and critically loaded systems
7
GI/G/1 Non-Stop Service
8
Queues in Tandem (with 1 bottleneck)
Bottleneck Server Just as simple…
9
Re-entrant Line bottleneck In the stable case:
10
Overloaded case --> Infinite Supply Re-entrant Line
Result:
11
Overloaded case --> Infinite Supply Re-entrant Line
1 6 8 1 2 3 5 6 4 8 7 9 Result:
12
Shocking result* coming up…
* at least for me
13
Back to Single Server (GI/G/1/K)
What happens here? Balancing Reduces Asymptotic Variance of Outputs Note: the figure assumes
14
BRAVO Effect (illustration for M/M/1)
More than a singular theoretic phenomenon
15
BRAVO Effect (for M/M/1/K)
16
Some (partial) intuition for M/M/1/K
Easy to see: 1 K K-1
17
Questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.