5 Lecture 5: Fluid Models
|
|
- Dwight Lewis Reeves
- 6 years ago
- Views:
Transcription
1 5 Lecture 5: Fluid Models Stability of fluid and stochastic processing networks Stability analysis of some fluid models Optimization of fluid networks. Separated continuous linear programming 5.1 Stability of fluid and stochastic processing networks The approach developed by Jim Dai and others (Harrison, Williams, Bramson, Rybko, Stolyar, Kumar, Meyn, Foss, Malyshev, Dumas) consists of the following steps: Primitives: Topology given by buffers k = 1,...,K, nodes i = 1,...,I and partition into constituencies C i, with constituency matrix C. Streams of external arrivals, services, and switches, E(t),S(t), Ψ(j 1,...,j K ), with SLLN rates, and more specifically with i.i.d. intervals. Policy All policies are work conserving and HL. specific ones include FIFO, SBP, GHLPS, GHLPPS. Traffic equations λ = α + P λ ρ = C(λ m) <e Queeuing network equations The standard queue balance equations, expected immediate workload and idle time process equations, and work conservation and head of the line conditions. In addition special equations for the policy. The Queeuing network equations and the initial state determine the Queueing network process X =(A(t),D(t),T(t),Y(t),W(t),Q(t)) Markov Process Additional information is necessary to obtain a Markovian state. e.g. for FIFO the information needs to include the class indices of the ordered customers at each node. The state space is a finite dimensional normed vector space. Fluid limits Use fluid scaling by parameter r of both time and space for X, with either fixed or r parametrized initial conditions. This yields (for bounded initial conditions as r ) a precompact family, which will have one or more Fluid Limits. Fluid model equations Under fluid scaling all the primitives have unique determitistic, continuous, linear fluid limits. Substituting these in the Queueing network equations gives the fluid model equations. The formal set of equations defines the fluid limit model. All its solution are called fluid model solutions. The main feature is that each fluid limit must be a fluid model solution. 1
2 Fluid model stability There are 4 major definitions: Stable will drive any Q() < 1toQ(t) = for t>δ. Weakly stable will keep Q(t) = for t>if Q() =. Unstable is not stable, at least one solution is at a sequance of times t n. Weakly unstable starting at Q() = every solution will pop up to Q(δ). WeaklyStable UnStable Stable WeaklyUnstable Figure 1: Stability or Instability of Fluid Models Theorems Stability or instability of the fluid model implies stability or instability of the original queueing network process as follows: Blow-Up If the fluid model is weakly unstable then the queueing network process will have Q(t) almost surely as t. Rate Stability Is defined by lim t D k (t)/t = λ k almost surely (note: this does not exclude Q(t) ). The queueing network process is rate stable if and only if the unique fluid limit for fixed initial data has Q(t) =. If the fluid model is weakly stable, the queueing network process is rate stable. If the fluid model is weakly un-stable, the queueing network process is not rate stable. Positive Harris recurrence If the fluid model is stable and the exogenous input intervals are unbounded and spread out, then the queueing network process is positive Harris recurrent. 5.2 Stability analysis of some fluid models For all fluid models, if s>t, T k (s) T k (t) <s t, hence all the components of X are Lipschitz continuous, hence absolutely continuous, hence are integrals of their derivatives which exist almost everywhere. We call t regular if derivatives exist. 2
3 Lemma 5.1 Let f : R + R + be absolutely conitnuous and ɛ>. If at every regular point t, f(t) > implies f(t) < ɛ, then f(t) =for all t>f()/ɛ. Proposition 5.2 (i) Q k (t) =implies Ȧk(t) =Ḋk(t). (ii) W i (t) > implies k C i m k Ḋ k (t) = 1 (iii) Ȧ(t) =α + P Ḋ(t). Lyapunov functions: Let g : R+ K R + be locally Lipschitz (Lipschitz on each compact set), and g(x) = only when x =. Let f(t) =g(q(t)). If for all regular t such that Q(t), f(t) < ɛ, then the fluid model is stable. g is called a Lyapunov function. Proposition 5.3 Areentrant line with ρ i < 1 is stable under LBFS Proof. Use f(t) =g(q(t)) = k Q k(t). Let k be last non-empty buffer, then f(t) =α 1 Ḋk(t) =α 1 1/ l C σ(k),l k m l α 1 ρ σ(k) 1 ρ σ(k) α 1 min i 1 ρ i ρ i Proposition 5.4 For a general fluid model, let λ be the solutions to the traffic equations, ɛ>. If (for regular t) Q k (t) > implies Ḋk(t) >λ k + ɛ, then the fluid model is stable. Proof. Use f(t) =e (I P ) 1 Q(t). This the sum over all buffers of the total amount of fluid anywhere in the system which needs to flow through each buffer. It satisfies: f(t) = f() + e λt e D(t). Let k be the buffers for which Q k (t) =, and for which therefore Q k (t) = A k (t) Ḋk(t) =. Let k c be the buffers for which Q k (t) >, and for which therefore, by assumption, Ḋ k c(t) > λ k c. We get: ( ) Ḋ k (t) = (I P k,k ) 1 α k + P k,k cḋkc(t) and therefore (I P k,k ( ) 1 α k + P k,k cλ k c(t)) = λ k f(t) = k k(λ k Ḋk(t)) + k k c (λ k Ḋk(t)) ɛ k c ɛ Corollary 5.5 For single class queueing networks (generalized Jackson networks), if ρ i < 1 the queuing network is positive Harris recurrent. 3
4 Proof. Take ɛ = min(µ k λ k ). Corollary 5.6 Multi class queueing networks with ρ i < 1, under the GHLPS policy with weights β k = λ k m k, are positive Harris recurrent. Proof. For GHLPS fluid model one has for Q k (t) > : T k (t) = β k l C σ(k),q l (t) β l 1 but in addition one also has for every fluid limit that T k (t) β k l C σ(k) β l. l C σ(k),q l (t)= Ȧ l (t) m l For β k = λ k m k this implies that if Q k (t) > then Ḋk(t) =λ k /ρ σ(k). Proposition 5.7 If for a subset of buffers B there are weights y k such that k B y kλ k m k > 1 and at the same time k B y k T k (t) 1. Then the fluid model is weakly unstable. Corollary 5.8 For single class queueing networks (generalized Jackson networks), if ρ i > 1 the fluid model is weakly unstable. Hence the queuing network workload diverges almost surely. Proof. Take B = C i,y k =1,i C i. Corollary 5.9 The Lu-Kumar network (machines i =1, 2; K =4, re-entrant path moves out, under SBP policy with high priority to k = 2, 4, is unstable if α 1 (m 2 + m 4 ) > 1. Proof. It can be shown that for the Lu Kumar queueing network, under the said SBP policy, if Q() =, then at all subsequent regular times t, T 2 (t) + T 4 (t) 1. This implies that the same holds for every fluid limit. Hence we can add this equation to the fluid model equations. Stability regions Stability region of a fluid model (under a given policy) is the region of system parameters (α, m > ) for which the fluid model is stable. Global stability region is the region of system parameters (α, m > ) for which the fluid model is stable under all work conserving head of the line policies. Theorem 5.1 (Bramson 96) Kelly queueing networks with ρ j < 1 under FIFO policy are stable. The proof uses an entropy of the whole path Lyapunov function +Wσ(k) (t) f(t) = k t where h(x) =x log x is the entropy function. 4 λ k h(ḋk(s)/λ k )ds
5 Piecewise linear Lyapunov functions Let L(t) =(I P ) 1 Q(t), it is the K-vector of total fluid anywhere in the system at time t, which needs to flow through each buffer. It satisfies: L k (t) =L k () + λ k t µ k T k (t),k = 1,...,K. Let y k > be some positive weights, and for queue length (fluid levels) vector q let h(y, q) =Cdiag(y)(I P ) 1 q. Let f i (t) = k C i y k L k (t) =h i (y, Q(t)), the weighted potential workload of machine i. Take f(t) = max i f i (t) as the Lyapunov function. Proposition 5.11 Suppose that for some ɛ> there exist weights y k such that: (i) W i (t) > implies f i (t) ɛ (ii) while f(t) > there exists f i (t) =f(t) for which W i (t) > (note, W i (t) is immediate work, while f i (t) is (weighted) potential work, hence W i (t) =does not imply f i (t) =). Then: Q(t) =for t f()/ɛ. These piecewise linear Lyapunov functions can be used to give a complete characterization of the global stability region for a two station multitype (multiclass with deterministic routing) fluid model. The following is a result leading to this characterization: Theorem 5.12 Atwo station fluid model is globally stable under ρ 1,ρ 2 < 1 if the parameters allow us to solve the following LP with objective value 1 (here e k is the kth unit vectoer): max s.t. ɛ l C σ(k) λ l y l y k µ k, k =1,...,K h 1 (y, e k ) h 2 (y, e k ), k C 2 h 2 (y, e k ) h 1 (y, e k ), k C 1 y k, k =1,...,K ɛ 1. Stabiliztion We saw that GHLPS with β k = λ k m k is stable (when all ρ i < 1). The policy of GRR, generalized round robin, where each machine serves it buffers cyclically, performing β k serivces at buffer k in each cycle (skipping to next buffer if buffer is empty before β k ), is stable for β k / l C σ(k) β l >m k λ k. Bramson has also shown that GHLPPS is stable. Another stabilizing mechanism is Leaky Bucket: Add K single class stations, one each in front of each buffer k of the original network, with service rate µ k = 2λ k 1+ρ σ(k). The resulting modified network is globally stable (when all ρ i < 1). Proof. The input load into station i of the original network is now ˆρ i = k C i µ k m k = 2 k C i λ k m k 1+ρ i = 2ρ i 1+ρ i < 1. One can then check that W i (t) > implies that Ẇi(t) =ˆρ i 1. So all the original stations will empty, and the leaky bucket stations will empty not long afterward, as they are single class stations with ρ<1. 5
6 5.3 Optimization of fluid networks Consider the queue balance equations, which for the fluid we write as Q(t) = Q() + A(t) D(t) = Q() + E(t) S(B(t)) + R (S (B(t)))1 Q(t) = Q() + A(t) D(t) = Q() + αt (I P )D(t) = Q() + αt (I P ) u(s)ds We shall switch notation, calling the fluid buffer levels x(t), and we let x() = Q() = a. Note also that u(s), the flow rates, are constrained by the processing capacities of the network. Single class fluid networks (I P ) u(s)ds + x(t) = a + αt Mu(s) 1 Here we have: x, u, a, α, m, µ are all I vectors, P is a substochastic I I matrix with spectral radius less than 1, and we have: (I P ) u(s)ds + x(t) = a + αt diag(m) u(s) 1 Here we let u(t) be an arbitrary control vector, the flow rates. For work conserving policies, we need to have u i (t) =µ i =1/m i as long as x i (t) >. Consider the problem of emptying the fluid network in minimal time. If the network is empty by time t, then: (I P ) u(s)ds = a + αt Or: Adding and subtracting diag(m)(i P ) 1 (a + αt) = diag(m) u(s)ds 1t diag(m)(i P ) 1 a ( 1 diag(m)(i P ) 1 α ) t We consider this component by component. Assume first that ( 1 diag(m)(i P ) 1 α ) i > (of course this is ρ i < 1). Then ( diag(m)(i P t T ) 1 a ) i = max T i T i = i (1 diag(m)(i P ) 1 α) i 6
7 Lemma 5.13 Consider Gu(s)ds + x(t) = a + αt, Hu(s) 1, x, u If x(t),u(t) are feasible trajectories between (t 1,t 2 ) with boundary states x(t 1 ),x(t 2 ), then the constant control ū =(t 2 t 1 ) 1 2 t 1 u(t)dt will give a new trajectory, ū(t) =ū, x(t) which is feasible, with x(t 1 )=x(t 1 ), x(t 2 )=x(t 2 ). We can empty the fluid network at time T, by using constant flow rates: u i (t) =(T i /T )µ i. Consider now the following trajectory: Start with all flows u i (t) =µ i, and continue until a buffer empties. For any buffer that is empty, keep the buffer empty, and reduce the value of u i. Continue until the next buffer empties, etc. The solution uses constant flow rates in the interval = t t 1 t I = T. This solution has the following characterizations: The trajectory is unique fluid solution for a work conserving policy. The trajectory is the oblique reflection of the (infeasible) trajectoty a + αt +(I P )µt (Harrison-Reiman). The trajectory solves the dynamic complementarity problem (Mandelbaum). The trajectory can be obtained by oblique projection of a + αt +(I P )µt to the positive orthant (Dupuis-Ramanan) The total fluid in the system is pathwise minimal for all feasible controls (Meyn). This trajectory is the solution of an SCLP (Weiss). If for some iρ i 1, we can define t T = max i:ρ i <1 T i T i = ( diag(m)(i P ) 1 a ) i (1 diag(m)(i P ) 1 α) i In that case all the non-bottleneck nodes can be emptied at (but not before) time T, and the flows out of all the remaining non-empty nodes is at rates µ i. Furthermore, if we construct the solution as before, we get piecewise constant flows in the intervals = t t 1 t l = T, and this solution has all the above properties. Single class fluid networks with multiple outflows Consider a single class queueing network in which at each node there are several servers available. List all these servers j =1,...,J, Server j will serve buffer i at node i, at a rate µ j, and upon completion of serivce by server j, customers from buffer i will move to l with probability 7
8 P j,l. We now have a new matrix G to replace I P, so that the jth column of G consists of δ i,l P i,l,l =1,...,I. The constraints on state and control are, with m j =1/µ j : Gu(s)ds + x(t) = a + αt diag(m) u(s) 1 x(t),u(t), t. To empty the system in minimum time, one has the linear program: min t GU = a + αt diag(m) U 1t U. Once this is solved, U are constant levels of flow which will empty the system at the minimum time T. Pursue the following policy: For U j = do not use server j. For all other j, use u j (t) =µ j as long as buffer i which j is serving is non-empty. You can now think of the various servers which you choose to utilize as being pooled to a single server with rate µ i = j:j serves i µ j and switching probabilities are P i,l = j:j serves i µ jp j,l / µ i. It can be shown that this policy minimizes the total fluid in the system path wise. The solution also has all the other properties above. 8
Operations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads
Operations Research Letters 37 (2009) 312 316 Contents lists available at ScienceDirect Operations Research Letters journal homepage: www.elsevier.com/locate/orl Instability of FIFO in a simple queueing
More informationOPEN MULTICLASS HL QUEUEING NETWORKS: PROGRESS AND SURPRISES OF THE PAST 15 YEARS. w 1. v 2. v 3. Ruth J. Williams University of California, San Diego
OPEN MULTICLASS HL QUEUEING NETWORKS: PROGRESS AND SURPRISES OF THE PAST 15 YEARS v 2 w3 v1 w 2 w 1 v 3 Ruth J. Williams University of California, San Diego 1 PERSPECTIVE MQN SPN Sufficient conditions
More informationMaximum pressure policies for stochastic processing networks
Maximum pressure policies for stochastic processing networks Jim Dai Joint work with Wuqin Lin at Northwestern Univ. The 2011 Lunteren Conference Jim Dai (Georgia Tech) MPPs January 18, 2011 1 / 55 Outline
More informationPositive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network. Haifa Statistics Seminar May 5, 2008
Positive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network Yoni Nazarathy Gideon Weiss Haifa Statistics Seminar May 5, 2008 1 Outline 1 Preview of Results 2 Introduction Queueing
More informationSTABILITY OF MULTICLASS QUEUEING NETWORKS UNDER LONGEST-QUEUE AND LONGEST-DOMINATING-QUEUE SCHEDULING
Applied Probability Trust (7 May 2015) STABILITY OF MULTICLASS QUEUEING NETWORKS UNDER LONGEST-QUEUE AND LONGEST-DOMINATING-QUEUE SCHEDULING RAMTIN PEDARSANI and JEAN WALRAND, University of California,
More informationGideon Weiss University of Haifa. Joint work with students: Anat Kopzon Yoni Nazarathy. Stanford University, MSE, February, 2009
Optimal Finite Horizon Control of Manufacturing Systems: Fluid Solution by SCLP (separated continuous LP) and Fluid Tracking using IVQs (infinite virtual queues) Stanford University, MSE, February, 29
More informationPositive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network
Positive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network Yoni Nazarathy a,1, Gideon Weiss a,1 a Department of Statistics, The University of Haifa, Mount Carmel 31905, Israel.
More informationStability and Asymptotic Optimality of h-maxweight Policies
Stability and Asymptotic Optimality of h-maxweight Policies - A. Rybko, 2006 Sean Meyn Department of Electrical and Computer Engineering University of Illinois & the Coordinated Science Laboratory NSF
More informationStability, Capacity, and Scheduling of Multiclass Queueing Networks
Stability, Capacity, and Scheduling of Multiclass Queueing Networks A THESIS Presented to The Academic Faculty by John Jay Hasenbein In Partial Fulfillment of the Requirements for the Degree of Doctor
More informationAdvanced Computer Networks Lecture 3. Models of Queuing
Advanced Computer Networks Lecture 3. Models of Queuing Husheng Li Min Kao Department of Electrical Engineering and Computer Science University of Tennessee, Knoxville Spring, 2016 1/13 Terminology of
More informationDynamic Scheduling of Multiclass Queueing Networks. Caiwei Li
Dynamic Scheduling of Multiclass Queueing Networks A Thesis Presented to The Academic Faculty by Caiwei Li In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Industrial
More informationOn Positive Harris Recurrence of Multiclass Queueing Networks: A Unified Approach Via Fluid Limit Models
On Positive Harris Recurrence of Multiclass Queueing Networks: A Unified Approach Via Fluid Limit Models J. G. Dai School of Industrial and Systems Engineering and School of Mathematics Georgia Institute
More informationMaximum Pressure Policies in Stochastic Processing Networks
OPERATIONS RESEARCH Vol. 53, No. 2, March April 2005, pp. 197 218 issn 0030-364X eissn 1526-5463 05 5302 0197 informs doi 10.1287/opre.1040.0170 2005 INFORMS Maximum Pressure Policies in Stochastic Processing
More information[4] T. I. Seidman, \\First Come First Serve" is Unstable!," tech. rep., University of Maryland Baltimore County, 1993.
[2] C. J. Chase and P. J. Ramadge, \On real-time scheduling policies for exible manufacturing systems," IEEE Trans. Automat. Control, vol. AC-37, pp. 491{496, April 1992. [3] S. H. Lu and P. R. Kumar,
More informationStability and Rare Events in Stochastic Models Sergey Foss Heriot-Watt University, Edinburgh and Institute of Mathematics, Novosibirsk
Stability and Rare Events in Stochastic Models Sergey Foss Heriot-Watt University, Edinburgh and Institute of Mathematics, Novosibirsk ANSAPW University of Queensland 8-11 July, 2013 1 Outline (I) Fluid
More information3 Lecture 3: Queueing Networks and their Fluid and Diffusion Approximation
3 Lecture 3: Queueing Networks and their Fluid and Diffusion Approximation Fluid and diffusion scales Queueing networks Fluid and diffusion approximation of Generalized Jackson networks. Reflected Brownian
More informationThis lecture is expanded from:
This lecture is expanded from: HIGH VOLUME JOB SHOP SCHEDULING AND MULTICLASS QUEUING NETWORKS WITH INFINITE VIRTUAL BUFFERS INFORMS, MIAMI Nov 2, 2001 Gideon Weiss Haifa University (visiting MS&E, Stanford)
More informationTransience of Multiclass Queueing Networks. via Fluid Limit Models. Sean P. Meyn. University of Illinois. Abstract
Transience of Multiclass Queueing Networks via Fluid Limit Models Sean P. Meyn University of Illinois Abstract This paper treats transience for queueing network models by considering an associated uid
More informationA Semiconductor Wafer
M O T I V A T I O N Semi Conductor Wafer Fabs A Semiconductor Wafer Clean Oxidation PhotoLithography Photoresist Strip Ion Implantation or metal deosition Fabrication of a single oxide layer Etching MS&E324,
More informationSTABILITY AND INSTABILITY OF A TWO-STATION QUEUEING NETWORK
The Annals of Applied Probability 2004, Vol. 14, No. 1, 326 377 Institute of Mathematical Statistics, 2004 STABILITY AND INSTABILITY OF A TWO-STATION QUEUEING NETWORK BY J. G. DAI, 1 JOHN J. HASENBEIN
More informationSTABILITY AND STRUCTURAL PROPERTIES OF STOCHASTIC STORAGE NETWORKS 1
STABILITY AND STRUCTURAL PROPERTIES OF STOCHASTIC STORAGE NETWORKS 1 by Offer Kella 2 and Ward Whitt 3 November 10, 1994 Revision: July 5, 1995 Journal of Applied Probability 33 (1996) 1169 1180 Abstract
More informationStability of queueing networks. Maury Bramson. University of Minnesota
Probability Surveys Vol. 5 (2008) 169 345 ISSN: 1549-5787 DOI: 10.1214/08-PS137 Stability of queueing networks Received June 2008. Maury Bramson University of Minnesota e-mail: bramson@math.umn.edu Contents
More informationStability and Heavy Traffic Limits for Queueing Networks
Maury Bramson University of Minnesota Stability and Heavy Traffic Limits for Queueing Networks May 15, 2006 Springer Berlin Heidelberg New York Hong Kong London Milan Paris Tokyo Contents 1 Introduction...............................................
More informationFLUID LIMITS TO ANALYZE LONG-TERM FLOW RATES OF A STOCHASTIC NETWORK WITH INGRESS DISCARDING
Submitted to the Annals of Applied Probability FLUID LIMITS TO ANALYZE LONG-TERM FLOW RATES OF A STOCHASTIC NETWORK WITH INGRESS DISCARDING By John Musacchio, and Jean Walrand, University of California,
More informationUNIQUE SOLVABILITY OF FLUID MODELS OF QUEUEING NETWORKS
UNIQUE SOLVABILITY OF FLUID MODELS OF QUEUEING NETWORKS I. NEDAIBORSHCH, K. NIKOLAEV AND A. VLADIMIROV Abstract. Deterministic fluid models imitate the long-term behavior of stochastic queueing networks.
More informationIntroduction to Markov Chains, Queuing Theory, and Network Performance
Introduction to Markov Chains, Queuing Theory, and Network Performance Marceau Coupechoux Telecom ParisTech, departement Informatique et Réseaux marceau.coupechoux@telecom-paristech.fr IT.2403 Modélisation
More informationState Space Collapse in Many-Server Diffusion Limits of Parallel Server Systems. J. G. Dai. Tolga Tezcan
State Space Collapse in Many-Server Diffusion imits of Parallel Server Systems J. G. Dai H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, Georgia
More informationFluid Models of Parallel Service Systems under FCFS
Fluid Models of Parallel Service Systems under FCFS Hanqin Zhang Business School, National University of Singapore Joint work with Yuval Nov and Gideon Weiss from The University of Haifa, Israel Queueing
More informationControl of multi-class queueing networks with infinite virtual queues
Control of multi-class queueing networks with infinite virtual queues Erjen Lefeber (TU/e) Workshop on Optimization, Scheduling and Queues Honoring Gideon Weiss on his Retirement June 8, 2012 Where innovation
More informationControl of Fork-Join Networks in Heavy-Traffic
in Heavy-Traffic Asaf Zviran Based on MSc work under the guidance of Rami Atar (Technion) and Avishai Mandelbaum (Technion) Industrial Engineering and Management Technion June 2010 Introduction Network
More informationPerformance Evaluation of Queuing Systems
Performance Evaluation of Queuing Systems Introduction to Queuing Systems System Performance Measures & Little s Law Equilibrium Solution of Birth-Death Processes Analysis of Single-Station Queuing Systems
More informationHEAVY-TRAFFIC LIMITS FOR STATIONARY NETWORK FLOWS. By Ward Whitt, and Wei You November 28, 2018
Stochastic Systems HEAVY-TRAFFIC LIMITS FOR STATIONARY NETWORK FLOWS By Ward Whitt, and Wei You November 28, 2018 We establish heavy-traffic limits for the stationary flows in generalized Jackson networks,
More informationDynamic Control of Parallel-Server Systems
Dynamic Control of Parallel-Server Systems Jim Dai Georgia Institute of Technology Tolga Tezcan University of Illinois at Urbana-Champaign May 13, 2009 Jim Dai (Georgia Tech) Many-Server Asymptotic Optimality
More informationConcave switching in single-hop and multihop networks
Queueing Syst (2015) 81:265 299 DOI 10.1007/s11134-015-9447-9 Concave switching in single-hop and multihop networks Neil Walton 1 Received: 21 July 2014 / Revised: 17 April 2015 / Published online: 23
More informationLinear Model Predictive Control for Queueing Networks in Manufacturing and Road Traffic
Linear Model Predictive Control for ueueing Networks in Manufacturing and Road Traffic Yoni Nazarathy Swinburne University of Technology, Melbourne. Joint work with: Erjen Lefeber (manufacturing), Hai
More informationMODELING WEBCHAT SERVICE CENTER WITH MANY LPS SERVERS
MODELING WEBCHAT SERVICE CENTER WITH MANY LPS SERVERS Jiheng Zhang Oct 26, 211 Model and Motivation Server Pool with multiple LPS servers LPS Server K Arrival Buffer. Model and Motivation Server Pool with
More informationISSN MPS-misc
J. G. Jim" Dai Stability of Fluid and Stochastic Processing Networks ISSN 1398-5981 MPS-misc 1999-9 Copyright (1998) by Jim Dai January 3, 1999 Web Sites http://www.isye.gatech.edu/faculty/dai/ Preface
More informationManagement of demand-driven production systems
Management of demand-driven production systems Mike Chen, Richard Dubrawski, and Sean Meyn November 4, 22 Abstract Control-synthesis techniques are developed for demand-driven production systems. The resulting
More informationIntro Refresher Reversibility Open networks Closed networks Multiclass networks Other networks. Queuing Networks. Florence Perronnin
Queuing Networks Florence Perronnin Polytech Grenoble - UGA March 23, 27 F. Perronnin (UGA) Queuing Networks March 23, 27 / 46 Outline Introduction to Queuing Networks 2 Refresher: M/M/ queue 3 Reversibility
More informationTHROUGHPUT ANALYSIS OF STOCHASTIC NETWORKS USING FLUID LIMITS. By John Musacchio University of California, Santa Cruz and
Submitted to the Annals of Applied Probability THROUGHPUT ANALYSIS OF STOCHASTIC NETWORKS USING FLUID LIMITS. By John Musacchio University of California, Santa Cruz and By Jean Walrand University of California,
More informationarxiv: v1 [math.pr] 13 Nov 2015
Submitted to the Annals of Applied Probability ON THE INSTABILITY OF MATCHING QUEUES arxiv:1511.04282v1 [math.pr] 13 Nov 2015 By Pascal Moyal and Ohad Perry Université de Technologie de Compiègne and Northwestern
More informationA tutorial on some new methods for. performance evaluation of queueing networks. P. R. Kumar. Coordinated Science Laboratory. University of Illinois
A tutorial on some new methods for performance evaluation of queueing networks P. R. Kumar Dept. of Electrical and Computer Engineering, and Coordinated Science Laboratory University of Illinois 1308 West
More informationarxiv: v2 [math.pr] 14 Feb 2017
Large-scale Join-Idle-Queue system with general service times arxiv:1605.05968v2 [math.pr] 14 Feb 2017 Sergey Foss Heriot-Watt University EH14 4AS Edinburgh, UK and Novosibirsk State University s.foss@hw.ac.uk
More informationStability in a 2-Station Re-Entrant Line under a Static Buffer Priority Policy and the Role of Cross-Docking in the Semiconductor Industry
Stability in a 2-Station Re-Entrant Line under a Static Buffer Priority Policy and the Role of Cross-Docking in the Semiconductor Industry A Thesis Presented to The Academic Faculty by Jozo Acksteiner
More informationLecture 9: Deterministic Fluid Models and Many-Server Heavy-Traffic Limits. IEOR 4615: Service Engineering Professor Whitt February 19, 2015
Lecture 9: Deterministic Fluid Models and Many-Server Heavy-Traffic Limits IEOR 4615: Service Engineering Professor Whitt February 19, 2015 Outline Deterministic Fluid Models Directly From Data: Cumulative
More informationarxiv: v1 [cs.sy] 10 Apr 2014
Concave Switching in Single and Multihop Networks arxiv:1404.2725v1 [cs.sy] 10 Apr 2014 Neil Walton 1 1 University of Amsterdam, n.s.walton@uva.nl Abstract Switched queueing networks model wireless networks,
More informationDynamic Matching Models
Dynamic Matching Models Ana Bušić Inria Paris - Rocquencourt CS Department of École normale supérieure joint work with Varun Gupta, Jean Mairesse and Sean Meyn 3rd Workshop on Cognition and Control January
More informationPiecewise Linear Test. Functions for Stability and Instability. of Queueing Networks. D. Down and S.P. Meyn. Abstract
Piecewise Linear Test Functions for Stability and Instability of Queueing Networks D. Down and S.P. Meyn Coordinated Science Laboratory 1308 W. Main Urbana, IL 61801 Abstract We develop the use of piecewise
More informationarxiv:math/ v4 [math.pr] 12 Apr 2007
arxiv:math/612224v4 [math.pr] 12 Apr 27 LARGE CLOSED QUEUEING NETWORKS IN SEMI-MARKOV ENVIRONMENT AND ITS APPLICATION VYACHESLAV M. ABRAMOV Abstract. The paper studies closed queueing networks containing
More informationExact Simulation of the Stationary Distribution of M/G/c Queues
1/36 Exact Simulation of the Stationary Distribution of M/G/c Queues Professor Karl Sigman Columbia University New York City USA Conference in Honor of Søren Asmussen Monday, August 1, 2011 Sandbjerg Estate
More informationAsymptotic Coupling of an SPDE, with Applications to Many-Server Queues
Asymptotic Coupling of an SPDE, with Applications to Many-Server Queues Mohammadreza Aghajani joint work with Kavita Ramanan Brown University March 2014 Mohammadreza Aghajanijoint work Asymptotic with
More informationQueueing Theory I Summary! Little s Law! Queueing System Notation! Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K "
Queueing Theory I Summary Little s Law Queueing System Notation Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K " Little s Law a(t): the process that counts the number of arrivals
More informationTwo Workload Properties for Brownian Networks
Two Workload Properties for Brownian Networks M. Bramson School of Mathematics University of Minnesota Minneapolis MN 55455 bramson@math.umn.edu R. J. Williams Department of Mathematics University of California,
More informationDesign, staffing and control of large service systems: The case of a single customer class and multiple server types. April 2, 2004 DRAFT.
Design, staffing and control of large service systems: The case of a single customer class and multiple server types. Mor Armony 1 Avishai Mandelbaum 2 April 2, 2004 DRAFT Abstract Motivated by modern
More informationTutorial: Optimal Control of Queueing Networks
Department of Mathematics Tutorial: Optimal Control of Queueing Networks Mike Veatch Presented at INFORMS Austin November 7, 2010 1 Overview Network models MDP formulations: features, efficient formulations
More informationSTABILITY OF TWO FAMILIES OF REAL-TIME QUEUEING NETWORKS ŁUKASZ K RU K (WARSZAWA)
PROBABILITY AND MATHEMATICAL STATISTICS Vol. 28, Fasc. 2 2008), pp. 79 202 STABILITY OF TWO FAMILIES OF REAL-TIME QUEUEING NETWORKS BY ŁUKASZ K RU K WARSZAWA) Abstract. We study open multiclass queueing
More informationA Diffusion Approximation for Stationary Distribution of Many-Server Queueing System In Halfin-Whitt Regime
A Diffusion Approximation for Stationary Distribution of Many-Server Queueing System In Halfin-Whitt Regime Mohammadreza Aghajani joint work with Kavita Ramanan Brown University APS Conference, Istanbul,
More informationSTABILITY OF JACKSON-TYPE QUEUEING NETWORKS, I
STABILITY OF JACKSON-TYPE QUEUEING NETWORKS, I François Baccelli INRIA Centre Sophia Antipolis 06565 Valbonne France; baccelli@sophia.inria.fr Serguei Foss NOVOSIBIRSK State University, 630090 Novosibirsk,
More informationA Push Pull Network with Infinite Supply of Work
DOI 10.1007/s11134-009-9121-1 A Push Pull Network with Infinite Supply of Work Anat Kopzon Yoni Nazarathy Gideon Weiss Received: 31 January 2007 / Revised: 28 July 2008 Springer Science+Business Media,
More informationScheduling Multiclass Queueing Networks via Fluid Models
Scheduling Multiclass Queueing Networks via Fluid Models John Hasenbein OR/IE, UT-Austin Ron Billings OR/IE, UT-Austin Leon Lasdon MSIS, UT-Austin Gideon Weiss Statistics, Univ of Haifa www.me.utexas.edu/~has
More informationFair Scheduling in Input-Queued Switches under Inadmissible Traffic
Fair Scheduling in Input-Queued Switches under Inadmissible Traffic Neha Kumar, Rong Pan, Devavrat Shah Departments of EE & CS Stanford University {nehak, rong, devavrat@stanford.edu Abstract In recent
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 22 12/09/2013. Skorokhod Mapping Theorem. Reflected Brownian Motion
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.7J Fall 213 Lecture 22 12/9/213 Skorokhod Mapping Theorem. Reflected Brownian Motion Content. 1. G/G/1 queueing system 2. One dimensional reflection mapping
More informationOn Dynamic Scheduling of a Parallel Server System with Partial Pooling
On Dynamic Scheduling of a Parallel Server System with Partial Pooling V. Pesic and R. J. Williams Department of Mathematics University of California, San Diego 9500 Gilman Drive La Jolla CA 92093-0112
More informationRouting and Staffing in Large-Scale Service Systems: The Case of Homogeneous Impatient Customers and Heterogeneous Servers 1
Routing and Staffing in Large-Scale Service Systems: The Case of Homogeneous Impatient Customers and Heterogeneous Servers 1 Mor Armony 2 Avishai Mandelbaum 3 June 25, 2008 Abstract Motivated by call centers,
More informationDirectional Derivatives of Oblique Reflection Maps
Directional Derivatives of Oblique Reflection Maps Avi Mandelbaum School of Industrial Engineering and Management, Technion, Haifa, Israel email: avim@techunixtechnionacil Kavita Ramanan Division of Applied
More informationA Mechanism for Pricing Service Guarantees
A Mechanism for Pricing Service Guarantees Bruce Hajek Department of Electrical and Computer Engineering and the Coordinated Science Laboratory University of Illinois at Urbana-Champaign Sichao Yang Qualcomm
More informationQueueing Networks and Insensitivity
Lukáš Adam 29. 10. 2012 1 / 40 Table of contents 1 Jackson networks 2 Insensitivity in Erlang s Loss System 3 Quasi-Reversibility and Single-Node Symmetric Queues 4 Quasi-Reversibility in Networks 5 The
More informationA Simple Memoryless Proof of the Capacity of the Exponential Server Timing Channel
A Simple Memoryless Proof of the Capacity of the Exponential Server iming Channel odd P. Coleman ECE Department Coordinated Science Laboratory University of Illinois colemant@illinois.edu Abstract his
More informationCollaboration and Multitasking in Networks: Architectures, Bottlenecks and Capacity
MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol. 00, No. 0, Xxxxx 0000, pp. 000 000 issn 1523-4614 eissn 1526-5498 00 0000 0001 INFORMS doi 10.1287/xxxx.0000.0000 c 0000 INFORMS Collaboration and Multitasking
More informationThroughput-optimal Scheduling in Multi-hop Wireless Networks without Per-flow Information
Throughput-optimal Scheduling in Multi-hop Wireless Networks without Per-flow Information Bo Ji, Student Member, IEEE, Changhee Joo, Member, IEEE, and Ness B. Shroff, Fellow, IEEE Abstract In this paper,
More informationMarkov Processes and Queues
MIT 2.853/2.854 Introduction to Manufacturing Systems Markov Processes and Queues Stanley B. Gershwin Laboratory for Manufacturing and Productivity Massachusetts Institute of Technology Markov Processes
More informationSTABILITY OF QUEUEING NETWORKS AND SCHEDULING POLICIES. P. R. Kumar and Sean P. Meyn y. Abstract
STABILITY OF QUEUEING NETWORKS AND SCHEDULING POLICIES P. R. Kumar and Sean P. Meyn y Abstract Usually, the stability of queueing networks is established by explicitly determining the invariant distribution.
More informationTales of Time Scales. Ward Whitt AT&T Labs Research Florham Park, NJ
Tales of Time Scales Ward Whitt AT&T Labs Research Florham Park, NJ New Book Stochastic-Process Limits An Introduction to Stochastic-Process Limits and Their Application to Queues Springer 2001 I won t
More informationOPTIMAL CONTROL OF STOCHASTIC NETWORKS - AN APPROACH VIA FLUID MODELS
OPTIMAL CONTROL OF STOCHASTIC NETWORKS - AN APPROACH VIA FLUID MODELS Nicole Bäuerle Department of Mathematics VII, University of Ulm D-8969 Ulm, Germany, baeuerle@mathematik.uni-ulm.de Abstract We consider
More informationMarkov processes and queueing networks
Inria September 22, 2015 Outline Poisson processes Markov jump processes Some queueing networks The Poisson distribution (Siméon-Denis Poisson, 1781-1840) { } e λ λ n n! As prevalent as Gaussian distribution
More informationSupply chains models
Supply chains models Maria Ivanova Technische Universität München, Department of Mathematics, Haupt Seminar 2016 Garching, Germany July 3, 2016 Overview Introduction What is Supply chain? Traffic flow
More informationTransitory Queueing Networks
OPERATIONS RESEARCH Vol. 00, No. 0, Xxxxx 0000, pp. 000 000 issn 0030-364X eissn 1526-5463 00 0000 0001 INFORMS doi 10.1287/xxxx.0000.0000 c 0000 INFORMS Authors are encouraged to submit new papers to
More informationA Heavy Traffic Approximation for Queues with Restricted Customer-Server Matchings
A Heavy Traffic Approximation for Queues with Restricted Customer-Server Matchings (Working Paper #OM-007-4, Stern School Business) René A. Caldentey Edward H. Kaplan Abstract We consider a queueing system
More informationHeavy-Traffic Optimality of a Stochastic Network under Utility-Maximizing Resource Allocation
Heavy-Traffic Optimality of a Stochastic Network under Utility-Maximizing Resource Allocation Heng-Qing Ye Dept of Decision Science, School of Business National University of Singapore, Singapore David
More informationCDA5530: Performance Models of Computers and Networks. Chapter 4: Elementary Queuing Theory
CDA5530: Performance Models of Computers and Networks Chapter 4: Elementary Queuing Theory Definition Queuing system: a buffer (waiting room), service facility (one or more servers) a scheduling policy
More informationIEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 3, MARCH
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 3, MARCH 1998 315 Asymptotic Buffer Overflow Probabilities in Multiclass Multiplexers: An Optimal Control Approach Dimitris Bertsimas, Ioannis Ch. Paschalidis,
More informationWorkload Models for Stochastic Networks: Value Functions and Performance Evaluation
Workload Models for Stochastic Networks: Value Functions and Performance Evaluation Sean Meyn Fellow, IEEE Abstract This paper concerns control and performance evaluation for stochastic network models.
More informationSTABILIZATION OF AN OVERLOADED QUEUEING NETWORK USING MEASUREMENT-BASED ADMISSION CONTROL
First published in Journal of Applied Probability 43(1) c 2006 Applied Probability Trust STABILIZATION OF AN OVERLOADED QUEUEING NETWORK USING MEASUREMENT-BASED ADMISSION CONTROL LASSE LESKELÄ, Helsinki
More informationThe G t /GI/s t + GI Many-Server Fluid Queue: Longer Online Version with Appendix
Queueing Systems manuscript No. (will be inserted by the editor) January 4, 22 The G t /GI/s t + GI Many-Server Fluid Queue: Longer Online Version with Appendix Yunan Liu Ward Whitt Received: date / Accepted:
More informationOn the static assignment to parallel servers
On the static assignment to parallel servers Ger Koole Vrije Universiteit Faculty of Mathematics and Computer Science De Boelelaan 1081a, 1081 HV Amsterdam The Netherlands Email: koole@cs.vu.nl, Url: www.cs.vu.nl/
More informationStability of the two queue system
Stability of the two queue system Iain M. MacPhee and Lisa J. Müller University of Durham Department of Mathematical Science Durham, DH1 3LE, UK (e-mail: i.m.macphee@durham.ac.uk, l.j.muller@durham.ac.uk)
More informationDELAY, MEMORY, AND MESSAGING TRADEOFFS IN DISTRIBUTED SERVICE SYSTEMS
DELAY, MEMORY, AND MESSAGING TRADEOFFS IN DISTRIBUTED SERVICE SYSTEMS By David Gamarnik, John N. Tsitsiklis and Martin Zubeldia Massachusetts Institute of Technology 5 th September, 2017 We consider the
More informationBy V. Pesic and R. J. Williams XR Trading and University of California, San Diego
Stochastic Systems 2016, Vol. 6, No. 1, 26 89 DOI: 10.1214/14-SSY163 DYNAMIC SCHEDULING FOR PARALLEL SERVER SYSTEMS IN HEAVY TRAFFIC: GRAPHICAL STRUCTURE, DECOUPLED WORKLOAD MATRIX AND SOME SUFFICIENT
More informationOther properties of M M 1
Other properties of M M 1 Přemysl Bejda premyslbejda@gmail.com 2012 Contents 1 Reflected Lévy Process 2 Time dependent properties of M M 1 3 Waiting times and queue disciplines in M M 1 Contents 1 Reflected
More informationarxiv:math/ v1 [math.pr] 5 Jul 2004
The Annals of Applied Probability 2004, Vol. 14, No. 3, 1055 1083 DOI: 10.1214/105051604000000224 c Institute of Mathematical Statistics, 2004 arxiv:math/0407057v1 [math.pr] 5 Jul 2004 FLUID MODEL FOR
More informationLoad Balancing in Distributed Service System: A Survey
Load Balancing in Distributed Service System: A Survey Xingyu Zhou The Ohio State University zhou.2055@osu.edu November 21, 2016 Xingyu Zhou (OSU) Load Balancing November 21, 2016 1 / 29 Introduction and
More informationStationary Probabilities of Markov Chains with Upper Hessenberg Transition Matrices
Stationary Probabilities of Marov Chains with Upper Hessenberg Transition Matrices Y. Quennel ZHAO Department of Mathematics and Statistics University of Winnipeg Winnipeg, Manitoba CANADA R3B 2E9 Susan
More informationConverse Lyapunov theorem and Input-to-State Stability
Converse Lyapunov theorem and Input-to-State Stability April 6, 2014 1 Converse Lyapunov theorem In the previous lecture, we have discussed few examples of nonlinear control systems and stability concepts
More informationOn the Asymptotic Optimality of the Gradient Scheduling Algorithm for Multiuser Throughput Allocation
OPERATIONS RESEARCH Vol. 53, No. 1, January February 2005, pp. 12 25 issn 0030-364X eissn 1526-5463 05 5301 0012 informs doi 10.1287/opre.1040.0156 2005 INFORMS On the Asymptotic Optimality of the Gradient
More informationApproximate Dynamic Programming for Networks: Fluid Models and Constraint Reduction
Approximate Dynamic Programming for Networks: Fluid Models and Constraint Reduction Michael H. Veatch Department of Mathematics Gordon College veatch@gordon.edu April 1, 005 Abstract This paper demonstrates
More informationQuiz 1 EE 549 Wednesday, Feb. 27, 2008
UNIVERSITY OF SOUTHERN CALIFORNIA, SPRING 2008 1 Quiz 1 EE 549 Wednesday, Feb. 27, 2008 INSTRUCTIONS This quiz lasts for 85 minutes. This quiz is closed book and closed notes. No Calculators or laptops
More informationNonlinear Control Systems
Nonlinear Control Systems António Pedro Aguiar pedro@isr.ist.utl.pt 3. Fundamental properties IST-DEEC PhD Course http://users.isr.ist.utl.pt/%7epedro/ncs2012/ 2012 1 Example Consider the system ẋ = f
More informationNEW FRONTIERS IN APPLIED PROBABILITY
J. Appl. Prob. Spec. Vol. 48A, 209 213 (2011) Applied Probability Trust 2011 NEW FRONTIERS IN APPLIED PROBABILITY A Festschrift for SØREN ASMUSSEN Edited by P. GLYNN, T. MIKOSCH and T. ROLSKI Part 4. Simulation
More informationStability, Queue Length and Delay of Deterministic and Stochastic Queueing Networks Cheng-Shang Chang IBM Research Division T.J. Watson Research Cente
Stability, Queue Length and Delay of Deterministic and Stochastic Queueing Networks Cheng-Shang Chang IBM Research Division T.J. Watson Research Center P.O. Box 704 Yorktown Heights, NY 10598 cschang@watson.ibm.com
More informationIn search of sensitivity in network optimization
In search of sensitivity in network optimization Mike Chen, Charuhas Pandit, and Sean Meyn June 13, 2002 Abstract This paper concerns policy synthesis in large queuing networks. The results provide answers
More information