Adviser: Frank, Yeong-Sung Lin Present by Wayne Hsiao.

Slides:



Advertisements
Similar presentations
Lect.3 Modeling in The Time Domain Basil Hamed
Advertisements

Hypothesis testing Another judgment method of sampling data.
Continuous-Time Markov Chains Nur Aini Masruroh. LOGO Introduction  A continuous-time Markov chain is a stochastic process having the Markovian property.
School of Information University of Michigan Network resilience Lecture 20.
Operations Research: Applications and Algorithms
Markov Chains.
1 A class of Generalized Stochastic Petri Nets for the performance Evaluation of Mulitprocessor Systems By M. Almone, G. Conte Presented by Yinglei Song.
Availability in Globally Distributed Storage Systems
1 Chapter 5 Continuous time Markov Chains Learning objectives : Introduce continuous time Markov Chain Model manufacturing systems using Markov Chain Able.
CS 795 – Spring  “Software Systems are increasingly Situated in dynamic, mission critical settings ◦ Operational profile is dynamic, and depends.
Operations Research: Applications and Algorithms
Introduction of Probabilistic Reasoning and Bayesian Networks
Markov Analysis Jørn Vatn NTNU.
Topics Review of DTMC Classification of states Economic analysis
Chapter 17 Markov Chains.
Reliable System Design 2011 by: Amir M. Rahmani
Lecture 13 – Continuous-Time Markov Chains
Introduction to Fault Diagnosis and Isolation(FDI) By Hariharan Kannan.
Introduction to PageRank Algorithm and Programming Assignment 1 CSC4170 Web Intelligence and Social Computing Tutorial 4 Tutor: Tom Chao Zhou
Network Resilience: Exploring Cascading Failures Vishal Misra Columbia University in the City of New York Joint work with Ed Coffman, Zihui Ge and Don.
A. BobbioBertinoro, March 10-14, Dependability Theory and Methods 5. Markov Models Andrea Bobbio Dipartimento di Informatica Università del Piemonte.
TCOM 501: Networking Theory & Fundamentals
Scheduling Algorithms for Wireless Ad-Hoc Sensor Networks Department of Electrical Engineering California Institute of Technology. [Cedric Florens, Robert.
Chapter 4: Stochastic Processes Poisson Processes and Markov Chains
1/55 EF 507 QUANTITATIVE METHODS FOR ECONOMICS AND FINANCE FALL 2008 Chapter 10 Hypothesis Testing.
Development of Empirical Models From Process Data
Condition State Transitions and Deterioration Models H. Scott Matthews March 10, 2003.
CHAPTER 6 Statistical Analysis of Experimental Data
Impact of Different Mobility Models on Connectivity Probability of a Wireless Ad Hoc Network Tatiana K. Madsen, Frank H.P. Fitzek, Ramjee Prasad [tatiana.
Lin Chen∗, Kaigui Bian∗, Lin Chen† Wei Yan∗, and Xiaoming Li∗
1 Spare part modelling – An introduction Jørn Vatn.
Queuing Networks: Burke’s Theorem, Kleinrock’s Approximation, and Jackson’s Theorem Wade Trappe.
The effect of New Links on Google Pagerank By Hui Xie Apr, 07.
3/25/2002Gustavo Cancelo1 Data flow analysis in the Processor Farmlet Transient Behavior of the M/M/1 process.
SIMULATION MODELING AND ANALYSIS WITH ARENA
Chapter 7. First and second order transient circuits
Isolated-Word Speech Recognition Using Hidden Markov Models
POWER CONTROL IN COGNITIVE RADIO SYSTEMS BASED ON SPECTRUM SENSING SIDE INFORMATION Karama Hamdi, Wei Zhang, and Khaled Ben Letaief The Hong Kong University.
Generalized Semi-Markov Processes (GSMP)
Random Sampling, Point Estimation and Maximum Likelihood.
Lecture 14 – Queuing Networks Topics Description of Jackson networks Equations for computing internal arrival rates Examples: computation center, job shop.
A review of M. Zonoozi, P. Dassanayake, “User Mobility and Characterization of Mobility Patterns”, IEEE J. on Sel. Areas in Comm., vol 15, no. 7, Sept.
Modelling and channel borrowing in mobile communications networks Richard J. Boucherie University of Twente Faculty of Mathematical Sciences University.
Network Survivability Against Region Failure Signal Processing, Communications and Computing (ICSPCC), 2011 IEEE International Conference on Ran Li, Xiaoliang.
Queuing Theory Basic properties, Markovian models, Networks of queues, General service time distributions, Finite source models, Multiserver queues Chapter.
1 Let X represent a Binomial r.v as in (3-42). Then from (2-30) Since the binomial coefficient grows quite rapidly with n, it is difficult to compute (4-1)
© The McGraw-Hill Companies, 2005 TECHNOLOGICAL PROGRESS AND GROWTH: THE GENERAL SOLOW MODEL Chapter 5 – second lecture Introducing Advanced Macroeconomics:
The Logistic Growth SDE. Motivation  In population biology the logistic growth model is one of the simplest models of population dynamics.  To begin.
Chapter 7 Sampling Distributions Statistics for Business (Env) 1.
Robustness of complex networks with the local protection strategy against cascading failures Jianwei Wang Adviser: Frank,Yeong-Sung Lin Present by Wayne.
Lecture 4: State-Based Methods CS 7040 Trustworthy System Design, Implementation, and Analysis Spring 2015, Dr. Rozier Adapted from slides by WHS at UIUC.
Generalized Semi- Markov Processes (GSMP). Summary Some Definitions The Poisson Process Properties of the Poisson Process  Interarrival times  Memoryless.
April 28, 2003 Early Fault Detection and Failure Prediction in Large Software Systems Felix Salfner and Miroslaw Malek Department of Computer Science Humboldt.
Chapter 61 Continuous Time Markov Chains Birth and Death Processes,Transition Probability Function, Kolmogorov Equations, Limiting Probabilities, Uniformization.
CS433 Modeling and Simulation Lecture 06 – Part 02 Discrete Markov Chains Dr. Anis Koubâa 11 Nov 2008 Al-Imam Mohammad.
Mitigation strategies on scale-free networks against cascading failures Jianwei Wang Adviser: Frank,Yeong-Sung Lin Present by Chris Chang.
1 8. One Function of Two Random Variables Given two random variables X and Y and a function g(x,y), we form a new random variable Z as Given the joint.
CS433 Modeling and Simulation Lecture 07 – Part 01 Continuous Markov Chains Dr. Anis Koubâa 14 Dec 2008 Al-Imam.
Yanjmaa Jutmaan  Department of Applied mathematics Some mathematical models related to physics and biology.
The generalization of Bayes for continuous densities is that we have some density f(y|  ) where y and  are vectors of data and parameters with  being.
STA347 - week 91 Random Vectors and Matrices A random vector is a vector whose elements are random variables. The collective behavior of a p x 1 random.
CS433 Modeling and Simulation Lecture 11 Continuous Markov Chains Dr. Anis Koubâa 01 May 2009 Al-Imam Mohammad Ibn Saud University.
A Framework for Network Survivability Characterization Soung C. Liew and Kevin W. Lu IEEE Journal on Selected Areas in Communications, January 1994 (ICC,
1 Chapter 5 Continuous time Markov Chains Learning objectives : Introduce continuous time Markov Chain Model manufacturing systems using Markov Chain Able.
O PTIMAL R EPLACEMENT AND P ROTECTION S TRATEGY FOR P ARALLEL S YSTEMS R UI P ENG, G REGORY L EVITIN, M IN X IE AND S ZU H UI N G Adviser: Frank, Yeong-Sung.
Discrete-time Markov chain (DTMC) State space distribution
Availability Availability - A(t)
V5 Stochastic Processes
5.1 Introduction to Curve Fitting why do we fit data to a function?
Chapter 9 Hypothesis Testing: Single Population
Presentation transcript:

Adviser: Frank, Yeong-Sung Lin Present by Wayne Hsiao

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 2

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 3

 Telecommunication networks have become one of the critical infrastructures  It is critically important that the network is survivable  The ability of the network to deliver the required services in the face of various disastrous events  Disaster propagation is one of the most common characteristics of disastrous events and has serious impact on communication networks 4

 Disaster propagation  A dynamic area-based event, in which the affected area can evolve spatially and temporally  For example, the 2005 hurricane Katrina in Louisiana, caused approximately 8% of all customarily routed networks in Louisiana outraged  The March 2011 earthquake and tsunami in east Japan, which cascaded from the center to Tohoku and Tokyo areas, damaged 1.9 million fixed-lines and 29 thousand wireless base stations 5

 Network design and operation need to consider survivability  This requires an understanding of the dynamical network recovery behaviors under failure patterns  To analyze the impact of disasters on the network as well as for estimating the benefits of alternative network survivable proposals, many mathematical models have been considered  However, up to now no much is known about the network survivability in the propagation of disastrous events 6

 The present paper develops a network survivability modeling method, which takes into consideration the propagating dynamics of disastrous events  The analysis is exemplified for three repair strategies.  The results not only are helpful in estimating quantitatively the survivability, but also provide insights on choosing among different repair strategies 7

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 8

 We focus on survivability as the ability of a networked system to continuously deliver services in compliance with the given requirements in the presence of failures and other undesired events  Network survivability is quantified as the transient performance from the instant when an undesirable event occurs until steady state with an acceptable performance level is attained  defined by the ANSI T1A1.2 committee 9

 The measure of interest M has the value m 0 before a failure occurs.  m a is the value of M just after the failure occurs  m u is the maximum difference between the value of M and m a after the failure  m r is the restored value of M after some time t r  t R is the relaxation time for the system to restore the value of M 10

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 11

 Develop such a model particularly for networked systems where disastrous events may propagate across geographical areas  A network can be viewed as a directed graph consisting of nodes and directed edges  Nodes represent the network infrastructures  The directed edges denote the directions of transitions  The network is vulnerable to all sorts of disaster, which may start on some network nodes and propagate to other nodes during a random time 12

 Suppose the number of nodes in the networked system is n  We consider a disastrous event, which occurs on these nodes in successive steps  The propagation is assumed to have ’memoryless’ property  The probability of disastrous events spreading from one given node to another depends only on the current system state but not on the history of the system  The affected node can be repaired (or replaced by a new one) in a random period  All the times of the disaster propagation and repair are exponentially distributed 13

 The state of each node of the system at time t lies within the set {0, 1}  At the initial time t = 0, a disastrous event affects the 1-st node and the system is in the state (0, 1,..., 1)  The disaster propagates from the node i − 1 to node i according to Poisson processes with rate λ i  A disastrous event can occur on only one node at a time  Each node has a specific repair process which is all at once and the repair time period of node i is exponentially distributed with mean value μ i 14

 The state of the system at any time t can be completely described by the collection of the state of each node  Where X i (t) = 0 (1 ≦ i ≦ n) if the event has occurred on the i-th node at time t, X i (t) = 1 in the case when the event has not occurred on the i-th node at time t. 15

 With the above assumptions, the transient process X(t) can be mathematically modeled as a continuous-time Markov chain (CTMC) with state space Ω = {(X 1, · · ·,X n ) : X 1, · · ·,X n ∈ {0, 1}}  The state space Ω consists of total N = 2 n states  The process X(t) starts in the state (0, 1,..., 1) and finishes in the absorbing state (1, 1,..., 1) 16

 Suppose that the system states are ordered so that in states 1, 2,...,N f (N f < N) the system has failure propagation and in states N f +1,N f +2,...,N the system is only in restoration phase  Then, the transition rate matrix Q = [qij] of the process {X(t), t ≧ 0} can be written in partitioned form as  where q ij denotes the rate of transition from state i to state j 17

 Let π (t) = { π i (t), i ∈ Ω } denote a row vector of transient state probabilities of X(t) at time t  With Q, the dynamic behavior of the CTMC can be described by the Kolmogorov differential-difference equation  Then the transient state probability vector can be obtained 18

NETWORK SURVIVABILITY UNDER DISASTER PROPAGATION (CONT.)  Let Υ i be the reward rate associated with state i  In our model, the performance is considered as reward  The network survivability performance is measured by the expected instantaneous reward rate E[M(t)] as 19

 An infrastructure wireless network example 20

 The state space of the chain is defined as S = {S 0,..., S Φ } ( Φ = 2 3 − 1)  State is described by a triple as (X 1, X 2, X 3 )  X i ∈ {0, 1} refers to the affected state of cell i, i = 1, 2, 3  The set of possible states is 21

 Two repair strategies  Scheme 1: each cell has its own repair facility  Scheme 2: all cells share a single repair facility 22

 Each cell i has its own repair facility with repair rate μ i  Fig. 3 shows the 8-state transition diagram of the CTMC model of the network example  The transition matrix is of size 8 × 8 and the initial probability vector is π = (1,0,0,0,0,0,0,0) 23

 Given a disaster occurs and destroys BS1, then all the users in cell 1 disconnect to the network  The initial state is (0, 1, 1)  The transition to state (0, 0, 1) occurs with rate λ 2 and takes into account the impact of disaster propagation from cell 1 to cell 2  The CTMC may also jump to original normal state (1, 1, 1) with repair rate μ 1 24

 On state (0, 0, 1), the CTMC may jump to three possible states  it may jump back to state (0, 1, 1) if the BS2 is repaired (this occurs with rate μ 2 )  it may jump to state (1, 0, 1) if the BS1 is repaired (this occurs with rate μ 1 )  the CTMC may jump to state (0,0,0) if the disaster propagates to cell 3 (this occurs with rate λ 3 ) 25

 Let π (t) = [ π (0,0,0) (t) · · · π (X1,X2,X3) (t) · · · π (1,1,1) (t)] denote the row vector of transient state probabilities at time t  The infinitesimal generator matrix for this CTMC is defined as Λ which is depicted in Fig. 4 26

 With Λ, the dynamic behavior of the CTMC can be described by the Kolmogorov differential- difference equation in the matrix form  π (t) can be solved using uniformization method  Let q ii be the diagnoal element of Λ and I be the unit matrix, then the transient state probability vector is obtained as follows: 27

SCHEME 1  Where β ≥ max i |q ii | is the uniform rate parameter and P = I+ Λ / β.  Truncate the summation to a large number (e.g., K), the controllable error ε can be computed from 28

 In the situation with this repair strategy, all cells share the same repair facility  The repair sequence is the same as the propagation path  cell1 → cell2 → cell3  The set of all possible states in this situation is: 29

 Accordingly, the transition diagram of the CTMC has the reduced 6- state as illustrated in Fig. 5 30

 The system is in each state k at time t, which is denoted by π k (t), k = 0,..., 5  They can be obtained in a closed-form by the convolution integration approach  Inserting Eq. (8) into Eq. (2) we can derive 31

 Continuing by induction, then we have 32

33

 We remark that simplification has been made in transition diagrams in Fig. 3 and Fig. 5  A cell which is recovered from a hurricane is unlikely to be destroyed by the same hurricane 34

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 35

 The expected instantaneous reward rate E[M(t)] gives the impact of users of the system at time t  Given the number of users N i of each cell i, as defined, the reward rate for each state is easily found 36

 The coverage radius of one BS is 1 km  For the three cells, we assume N 1 = 3000,N 2 = 5000, N 3 = 2000  For the setting of propagation rates, We refer to the data from Hurricane Katrina situation report  The peak wind speed was reported as high as 115 mph (184 km/h)  The units of repair time of BS is hours  It is acceptable that the disaster propagation rates are more than two order of magnitude than repair rates 37

 In Fig. 6, where the chosen repair strategy is Scheme 1  Consider the scenario  The fault propagation rate is high ( λ 2 = 5, λ 3 = 5), and the repair rates ( μ 1 = 0.04, μ 2 = 0.08, μ 3 = 0.12) are low  In this scenario, the fraction of active users is low (roughly 0.07, 2 hours after the failure)  If the repair rates are relatively higher ( μ 1 = 0.36, μ 2 = 0.72, μ 3 = 1.08), the fraction of active users sharply increase  The effect of the fault propagation rate is not as evident for longer observation time (after 10 hours) dd 38

 The plus-marked and dashed (blue) curves cross each other at time t ≈ 2 at Fig. 6  If we account for up to roughly two hours after the disaster, the fault propagation rates affect the service performance more than the repair rates  In contrast, if we account for longer periods of time, the repairs rates yield more benefits than to have lower fault propagation rate 39

 In the following, we compare three repair schemes  Scheme 1  Scheme 2  Scheme 3: same as Scheme 1 but with double repair rates 2 μ 1, 2 μ 2, 2 μ 3 40

NUMERICAL RESULT (CONT.) dd 41

 Introduction  Network Survivability  Network Survivability under Disaster Propagation  Numerical Result  Conclusion 42

 We have modeled the survivability of an infrastructure- based wireless network by a CTMC that incorporates the correlated failures caused by disaster propagation  The focus has been on computing the transient reward measures of the model  Numerical results have been presented to study the impact of the underlying parameters and different repair strategies on network survivability 43

44 Thanks for Your Listening !