## Fast Simulation of Packet Loss Rates in a Shared Buffer Communications Switch (2001)

Venue: | ACM Transactions on Modeling and Computer Simulation |

Citations: | 19 - 1 self |

### BibTeX

@ARTICLE{Chang01fastsimulation,

author = {Cheng-shang Chang and Philip Heidelberger and Perwez Shahabuddin},

title = {Fast Simulation of Packet Loss Rates in a Shared Buffer Communications Switch},

journal = {ACM Transactions on Modeling and Computer Simulation},

year = {2001},

volume = {5},

pages = {306--325}

}

### OpenURL

### Abstract

This paper describes an efficient technique for estimating, via simulation, the probability of buffer overows in a queueing model that arises in the analysis of ATM (Asynchronous Transfer Mode) communication switches. There are multiple streams of (autocorrelated) traffic feeding the switch that has a buffer of finite capacity. Each stream is designated as either being of high or low priority. When the queue length reaches a certain threshold, only high priority packets are admitted to the switch's buffer. The problem is to estimate the loss rate of high priority packets. An asymptotically optimal importance sampling approach is developed for this rare event simulation problem. In this approach, the importance sampling is done in two distinct phases. In the first phase, an importance sampling change of measure is used to bring the queue length up to the threshold at which low priority packets get rejected. In the second phase, a different importance sampling change of measure is used to move the queue length from the threshold to the buffer capacity.

### Citations

615 |
Applied Probability and Queues
- Asmussen
- 2003
(Show Context)
Citation Context ...ing and reliability theory is given in [26], while a survey on fast simulation of reliability models is given in [32]. Relevant references on the use of importance sampling in queueing models include =-=[1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]-=-. In this paper, we consider simulation of a single switch, the architecture and control of which has been proposed for use in ATM networks [15, 16]. The switch operates in discrete time. It is fed by... |

469 |
Monte Carlo Methods
- Hammersley, Handscomb
- 1964
(Show Context)
Citation Context ...ce of ATM models. Standard simulation of such rare events is known to take a prohibitive amount of time. Therefore, research has focused on how to apply a general technique called importance sampling =-=[23, 25]-=- so as to speed up the simulation of such rare events. A survey, with numerous references, on the application of importance sampling to rare event simulation of models in queueing and reliability theo... |

450 |
Equivalent capacity and its application to bandwidth allocation in high-speed networks
- Guerin, Ahmadi, et al.
(Show Context)
Citation Context ...is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [3, 6, 7, 13, 20, 30, 36]. The relationship between \eective bandwidths" (see, e.g=-=., [8, 14, 21, 24, 27, 39-=-]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals with the rate at which large queue length probabilities decay.) In this paper, we sho... |

196 |
Large Deviation Techniques in Decision, Simulation, and Estimation
- Bucklew
- 1990
(Show Context)
Citation Context ...ce of the estimate goes to zero at the fastest possible rate.) This change of measure is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [=-=3, 6, 7, 13, 20, 30, 36]. T-=-he relationship between \eective bandwidths" (see, e.g., [8, 14, 21, 24, 27, 39]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals w... |

191 | Effective bandwidths at multi-class queues
- Kelly
- 1991
(Show Context)
Citation Context ...is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [3, 6, 7, 13, 20, 30, 36]. The relationship between \eective bandwidths" (see, e.g=-=., [8, 14, 21, 24, 27, 39-=-]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals with the rate at which large queue length probabilities decay.) In this paper, we sho... |

172 |
Fast simulation of rare events in queueing and reliability models
- Heidelberger
- 1995
(Show Context)
Citation Context ... up the simulation of such rare events. A survey, with numerous references, on the application of importance sampling to rare event simulation of models in queueing and reliability theory is given in =-=[26]-=-, while a survey on fast simulation of reliability models is given in [32]. Relevant references on the use of importance sampling in queueing models include [1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 3... |

114 |
Importance sampling for stochastic simulations
- Glynn, Iglehart
- 1989
(Show Context)
Citation Context ...ce of ATM models. Standard simulation of such rare events is known to take a prohibitive amount of time. Therefore, research has focused on how to apply a general technique called importance sampling =-=[23, 25]-=- so as to speed up the simulation of such rare events. A survey, with numerous references, on the application of importance sampling to rare event simulation of models in queueing and reliability theo... |

99 |
Effective bandwidths for the multi-type UAS channel
- Gibbens, Hunt
- 1991
(Show Context)
Citation Context ...is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [3, 6, 7, 13, 20, 30, 36]. The relationship between \eective bandwidths" (see, e.g=-=., [8, 14, 21, 24, 27, 39-=-]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals with the rate at which large queue length probabilities decay.) In this paper, we sho... |

89 |
Tail probabilities with statistical multiplexing and effective bandwidths in multi-class queues
- Whitt
- 1993
(Show Context)
Citation Context |

75 |
A quick simulation method for excessive backlogs in networks of queues
- Parekh, Walrand
- 1989
(Show Context)
Citation Context ...ing and reliability theory is given in [26], while a survey on fast simulation of reliability models is given in [32]. Relevant references on the use of importance sampling in queueing models include =-=[1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]-=-. In this paper, we consider simulation of a single switch, the architecture and control of which has been proposed for use in ATM networks [15, 16]. The switch operates in discrete time. It is fed by... |

72 |
Probability
- Breiman
- 1968
(Show Context)
Citation Context ...e error is also obtained if d 1 f(B) L(X) Df(B) for all X 2 R B where D is a random variable such that E p 0 [D 2 ]s1. To estimate the steady state packet loss rate, r, we exploit a ratio formula [5=-=, 11]-=- that is an extension of the ratio formula that holds in regenerative systems [12, 38]. (While, with multiple Markovian sources the model may theoretically regenerative, the regenerations may occur so... |

71 |
Large deviations for a general class of random vectors
- Ellis
- 1984
(Show Context)
Citation Context ...ce of the estimate goes to zero at the fastest possible rate.) This change of measure is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [=-=3, 6, 7, 13, 20, 30, 36]. T-=-he relationship between \eective bandwidths" (see, e.g., [8, 14, 21, 24, 27, 39]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals w... |

55 | Effective Bandwidth and Fast Simulation of ATM Intree Networks
- Chang, I-Ieidelberger, et al.
- 1993
(Show Context)
Citation Context ...(Indeed, one application of this paper is to facilitate such an understanding.) Fast simulation of such a switch without priorities (i.e., all packets are high priority and B 2 = 0) was considered in =-=[9, 10]-=-. In those papers, the asymptotically optimal (as 1 B ! 1) importance sampling change of measure was described. (Here, asymptotically optimal means that the variance of the estimate goes to zero at th... |

53 |
A Guide to Simulation, Second Edition
- Bratley, Fox, et al.
- 1987
(Show Context)
Citation Context ...pair of A cycles. Then, the estimate of r is given by ^ r n ( ) = P n i=1 L i Y i P n i=1 i : (15) Because the A cycles are not independent and identically distributed, the method of batch means [4] can be used for variance estimation. 3 Fast Simulation of the Shared Buer Switch Let us now turn to the simulation of the shared buer switch. Let H denote the set of indices of high priority arriva... |

49 |
Importance sampling in the Monte Carlo study of sequential tests
- Siegmund
- 1976
(Show Context)
Citation Context ...ing and reliability theory is given in [26], while a survey on fast simulation of reliability models is given in [32]. Relevant references on the use of importance sampling in queueing models include =-=[1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]-=-. In this paper, we consider simulation of a single switch, the architecture and control of which has been proposed for use in ATM networks [15, 16]. The switch operates in discrete time. It is fed by... |

41 |
Risk theory in a Markovian environment
- Asmussen
- 1989
(Show Context)
Citation Context |

41 |
Statistical multiplexing with loss priorities in rate-based congestion control of high-speed networks
- Elwadi, Mitra
- 1994
(Show Context)
Citation Context ...ls include [1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]. In this paper, we consider simulation of a single switch, the architecture and control of which has been proposed for use in ATM networks =-=[15, 16-=-]. The switch operates in discrete time. It is fed by K (K 1) independent streams of trac. Each stream is designated as being either a high priority or low priority stream. The switch can service up ... |

33 |
Regenerative stochastic processes
- Smith
- 1955
(Show Context)
Citation Context ...om variable such that E p 0 [D 2 ]s1. To estimate the steady state packet loss rate, r, we exploit a ratio formula [5, 11] that is an extension of the ratio formula that holds in regenerative systems =-=[12, 38]-=-. (While, with multiple Markovian sources the model may theoretically regenerative, the regenerations may occur so infrequently as to make the regenerative method useless for practical purposes.) For ... |

32 |
Fluid Models for the Analysis and Design of Statistical Multiplexing with Loss Priorities on Multiple Classes of Bursty Traf lc: INFOCOM’92 W.K.Grassman, M.I.Taksar and D.P.Heyman Regeneratiwe Analysis and Steady State Distributions for Markov Chains: Ope
- Elwalid, Mitra
- 1985
(Show Context)
Citation Context ...ls include [1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]. In this paper, we consider simulation of a single switch, the architecture and control of which has been proposed for use in ATM networks =-=[15, 16-=-]. The switch operates in discrete time. It is fed by K (K 1) independent streams of trac. Each stream is designated as being either a high priority or low priority stream. The switch can service up ... |

26 |
Simulating stable stochastic systems, iii: Regenerative processes and discrete-event simulation
- Crane, Iglehart
- 1975
(Show Context)
Citation Context ...om variable such that E p 0 [D 2 ]s1. To estimate the steady state packet loss rate, r, we exploit a ratio formula [5, 11] that is an extension of the ratio formula that holds in regenerative systems =-=[12, 38]-=-. (While, with multiple Markovian sources the model may theoretically regenerative, the regenerations may occur so infrequently as to make the regenerative method useless for practical purposes.) For ... |

26 |
On the optimality and stability of exponential twisting in Monte Carlo estimation
- Sadowsky
- 1993
(Show Context)
Citation Context |

24 |
Monte Carlo simulation and large deviations theory for uniformly recurrent Markov chains
- Bucklew, Ney, et al.
- 1990
(Show Context)
Citation Context ...ce of the estimate goes to zero at the fastest possible rate.) This change of measure is closely related to the Gartner-Ellis theorem and the theory of large deviations for Markov additive processes [=-=3, 6, 7, 13, 20, 30, 36]. T-=-he relationship between \eective bandwidths" (see, e.g., [8, 14, 21, 24, 27, 39]) and asymptotically optimal fast simulation is also explored in [9, 10]. (The theory of eective bandwidths deals w... |

20 |
On large deviations from the invariant measure, Theory Probab
- Gärtner
- 1977
(Show Context)
Citation Context |

18 |
Conjugate process and the simulation of ruin problems
- Asmussen
(Show Context)
Citation Context |

18 |
queue length and delay, Part II: Stochastic queueing networks
- Chang, “Stability
- 1992
(Show Context)
Citation Context |

18 |
Simulating level crossing probabilities by importance sampling
- Lehtonen, Nyrhinen
- 1992
(Show Context)
Citation Context |

14 |
Eective bandwidth of general markovian traÆc sources and admission control of high speed networks
- Elwalid, Mitra
- 1993
(Show Context)
Citation Context |

12 | Fast simulation of steady-state availability in non-Markovian highly dependable systems
- Nicola, Shahabuddin, et al.
- 1993
(Show Context)
Citation Context ...that tosnd the optimal importance sampling change of measure requires solving Equation 12, a ( ) = c, a nonlinear equation involving the spectral radiuses of the sources. A \splitting" techniqu=-=e [9, 31-=-] can be used to eciently estimate both E[Y ] and E[]. The process is simulated (without importance sampling) until it is approximately in steady state. Then whenever the process enters A, two A-cycle... |

8 |
A uniform theory for sums of Markov chain transition probabilities. The Annals of Probability 3
- Cogburn
- 1975
(Show Context)
Citation Context ...e error is also obtained if d 1 f(B) L(X) Df(B) for all X 2 R B where D is a random variable such that E p 0 [D 2 ]s1. To estimate the steady state packet loss rate, r, we exploit a ratio formula [5=-=, 11]-=- that is an extension of the ratio formula that holds in regenerative systems [12, 38]. (While, with multiple Markovian sources the model may theoretically regenerative, the regenerations may occur so... |

6 |
Overflow probabilities in Jackson networks
- Glasserman, Kou
- 1993
(Show Context)
Citation Context |

4 |
Fast estimation of the statistics of excessive backlogs in tandem networks of queues. Australian Telecommun
- Frater, Anderson
- 1989
(Show Context)
Citation Context |

4 |
On large deviations theory and asymptotically ecient Monte Carlo estimation
- Sadowsky, Bucklew
- 1990
(Show Context)
Citation Context |

3 |
Optimally ecient estimation of the statistics of rare events in queueing networks
- Frater, Lennon, et al.
- 1991
(Show Context)
Citation Context |

3 |
Quick simulation of ATM buers with on-o multiclass Markov sources
- Kesidis, Walrand
- 1993
(Show Context)
Citation Context |

3 |
Techniques for fast simulation of highly dependable systems
- Nicola, Shahabuddin, et al.
- 1993
(Show Context)
Citation Context ..., on the application of importance sampling to rare event simulation of models in queueing and reliability theory is given in [26], while a survey on fast simulation of reliability models is given in =-=[32]-=-. Relevant references on the use of importance sampling in queueing models include [1, 2, 3, 17, 18, 19, 22, 28, 29, 33, 34, 35, 37]. In this paper, we consider simulation of a single switch, the arch... |

2 |
Large deviations and ecient simulation of excessive backlogs in a GI/G/m queue
- Sadowsky
- 1991
(Show Context)
Citation Context |

1 | The application of eective bandwidth to fast simulation of communication networks - Chang, Heidelberger, et al. - 1993 |

1 |
Rare events and reverse time models
- Frater, Bitmead, et al.
- 1989
(Show Context)
Citation Context |

1 |
On asymptotically ecient simulation of ruin probabilities in a Markovian environment
- Lehtonen, Nyrhinen
- 1992
(Show Context)
Citation Context |