## Queueing Delays in Rate Controlled ATM Networks (1993)

Venue: | In Proceedings of INFOCOM '93 |

Citations: | 12 - 2 self |

### BibTeX

@INPROCEEDINGS{Banerjea93queueingdelays,

author = {Anindo Banerjea and Srinivasan Keshav},

title = {Queueing Delays in Rate Controlled ATM Networks},

booktitle = {In Proceedings of INFOCOM '93},

year = {1993}

}

### OpenURL

### Abstract

This paper addresses the problem of finding the worst case end-to-end delay and buffer occupancy bounds in ATM networks with rate-controlled, non-work conserving servers. A theoretical framework is constructed to analyze such servers in isolation and in tandem. The analysis is based on a simple fluid model, but care is taken so that the computed delay and buffer occupancy values are upper bounds on actual values. A simple algorithm is presented to perform these calculations in linear time. Simulation results compare the computed worst case delays with the actual delays obtained on some simple network topologies. The algorithm is found to predict node delays well for bursty input traffic, but poorly for smooth input traffic. Buffer requirements are predicted well in both cases. 1. Introduction Recent work has shown that framed, non-workconserving servers can provide end-to-end delay bounds to users who need strict guarantees on network performance [15]. By framed we mean that the serv...

### Citations

1673 | A generalized processor sharing approach to flow control in integrated services networks ⎯ the single node case
- Parekh, Gallager
- 1993
(Show Context)
Citation Context ...ximum delay, but also for the delay distribution [10]. However, this work assumes a work-conserving service discipline, and a stronger characterization of input traffic than ours. Parekh and Gallager =-=[12, 13]-=- have computed worst case delays assuming sources were leaky bucket compliant and the servers used the Packetized Generalized Processor Sharing (PGPS) scheduling discipline. This is a workconserving d... |

631 | A scheme for real-time channel establishment in wide-area networks
- Ferrari, Verma
- 1990
(Show Context)
Citation Context ...hat allow us to present an algorithm to compute the worst case delay and buffer bounds. 3.1. Input Traffic Model We characterize the traffic entering the channel with three parameters as suggested in =-=[5, 6]-=-: x min The minimum cell inter-arrival time. x ave The minimum average cell inter-arrival time, over any time interval of length I. I The averaging interval for calculating x ave . Recent work has sho... |

557 |
A calculus for network delay, part I: Network elements in isolation
- Cruz
- 1991
(Show Context)
Citation Context ...st, we will assume parametrically constrained inputs and compute worst-case delays. This type of analysis was first proposed by Cruz who used it to analyze some work-conserving scheduling disciplines =-=[2, 3]-=-. In recent work, Cruz has also computed delay bounds for non-work conserving disciplines, but this analysis does not explicitly consider the effects of different frame sizes at different servers, res... |

300 |
A calculus for network delay, part II: Network analysis
- Cruz
- 1991
(Show Context)
Citation Context ...st, we will assume parametrically constrained inputs and compute worst-case delays. This type of analysis was first proposed by Cruz who used it to analyze some work-conserving scheduling disciplines =-=[2, 3]-=-. In recent work, Cruz has also computed delay bounds for non-work conserving disciplines, but this analysis does not explicitly consider the effects of different frame sizes at different servers, res... |

183 | Comparison of Rate-Based Service Disciplines
- Zhang, Keshav
- 1991
(Show Context)
Citation Context ...dicted well in both cases. 1. Introduction Recent work has shown that framed, non-workconservingsservers can provide end-to-end delay bounds to users who need strict guarantees on network performance =-=[15]-=-. By framed we mean that the servers work on the basis of a fixed size interval called the frametime, during which they allocate a number of transmission slots to each channel being served. We assume ... |

143 |
On computing per-session performance bounds in high-speed multi-hop computer networks
- Kurose
- 1992
(Show Context)
Citation Context ...effects of different frame sizes at different servers, resulting in weaker delay bounds [4]. Kurose has computed tight delay bounds not only for the maximum delay, but also for the delay distribution =-=[10]-=-. However, this work assumes a work-conserving service discipline, and a stronger characterization of input traffic than ours. Parekh and Gallager [12, 13] have computed worst case delays assuming sou... |

104 |
A Stop-and-Go Queueing Framework for Congestion Management
- Golestani
- 1990
(Show Context)
Citation Context ...send out the available cells, then use the free slots for non-real time traffic. Examples of such servers are those that obey the Hierarchical Round Robin (HRR) and the Stop-and-Go service discipline =-=[8, 7]-=-. These servers simultaneously serve a number of frames of different lengths so that a choice of frametimes is available during channel establishment. While earlier work considered the behavior of a s... |

80 | The REAL network simulator
- Keshav
- 1988
(Show Context)
Citation Context ... running simulations for long enough periods, we achieve convergence of the average delay, as described in Section 7.1. 7.1. Simulation Details All simulations were performed using the REAL simulator =-=[9]-=-. The simulation topology is simple - a series of switches from a source (labeled 1) to a destination (Figure 5). This source sends parametrically constrained greedy traffic, that is, traffic that can... |

56 |
Service burstiness and dynamic burstiness measures: A framework
- Cruz
- 1992
(Show Context)
Citation Context ...o computed delay bounds for non-work conserving disciplines, but this analysis does not explicitly consider the effects of different frame sizes at different servers, resulting in weaker delay bounds =-=[4]-=-. Kurose has computed tight delay bounds not only for the maximum delay, but also for the delay distribution [10]. However, this work assumes a work-conserving service discipline, and a stronger chara... |

30 |
Rate Controlled Servers for Very High Speed Networks
- Kalmanek, Kanakia, et al.
- 1990
(Show Context)
Citation Context ...send out the available cells, then use the free slots for non-real time traffic. Examples of such servers are those that obey the Hierarchical Round Robin (HRR) and the Stop-and-Go service discipline =-=[8, 7]-=-. These servers simultaneously serve a number of frames of different lengths so that a choice of frametimes is available during channel establishment. While earlier work considered the behavior of a s... |

28 |
A simple theory of traffic and resource allocation
- Low, Varaiya
- 1991
(Show Context)
Citation Context ...eir notion of `greedy' sources extensively in our work. Our formal modelling of a message stream and of a server as a transformation on a message stream is based on the recent work of Low and Varaiya =-=[11]-=-. While they use their model mainly to analyze burstiness, we use it for exact transient analysis of a tandem of rate-controlled servers. Our server model is based on the HRR servers described in [8].... |

6 |
Client Requirements for Real Time Communication Services
- Ferrari
- 1990
(Show Context)
Citation Context ...hat allow us to present an algorithm to compute the worst case delay and buffer bounds. 3.1. Input Traffic Model We characterize the traffic entering the channel with three parameters as suggested in =-=[5, 6]-=-: x min The minimum cell inter-arrival time. x ave The minimum average cell inter-arrival time, over any time interval of length I. I The averaging interval for calculating x ave . Recent work has sho... |

3 |
Guaranteed Performance Communication in High-Speed Networks
- Verma
- 1991
(Show Context)
Citation Context ...interval of length I. I The averaging interval for calculating x ave . Recent work has shown that this model accurately describes many types of rate-controlled traffic expected in high-speed networks =-=[14]-=-. We assume that the user of the channel is required to obey these restrictions on the input traffic, and the delay and buffer values calculated need only hold if these restrictions are not violated. ... |

1 |
A Formal Analysis of Queueing Delays
- Banerjea, Keshav
- 1992
(Show Context)
Citation Context ...nputs. We first look at a single server, then consider servers operating in tandem, and finally introduce, define and analyze servers with slippage. For reasons of space, we defer proofs to Reference =-=[1]-=-. 3.3.1. Single Server Lemma 1.1: If traffic generated by a PCS M in (t) is input to a fixed rate server with rate rs1/x ave , then any interval of time during which Q M in (t) > 0 continuously has to... |