Michael Rumsewicz

Wavelength converters in dynamically-reconfigurable WDM networks

In simple wavelength-division multiplexed (WDM) networks, a connection must be established along a route using a common wavelength on all of the links along the route. This constraint may be removed by the introduction of wavelength converters, which are devices which take the data modulated on an input wavelength and transfer it to a different output wavelength. Wavelength converters thus improve network blocking performance. However, the introduction of wavelength converters into WDM cross-connects increases the hardware cost and complexity. Thus, it is important to establish precisely what advantages wavelength converters offer WDM networks. There has been considerable interest in the literature in the performance improvements offered by the introduction of wavelength converters into dynamically-reconfigurable WDM networks. This article provides a review of the conclusions drawn from these investigations. The performance improvements offered by wavelength converters depend on a number of factors, including network topology and size, the number of wavelengths, and the routing and wavelength assignment algorithms used. We discuss these factors here. However, it has been shown that wavelength converters offer only modest performance improvements in many networks. We also consider networks with limited wavelength conversion, in which the set of allowable conversions at a network node is constrained by having limited numbers of wavelength converters, or by using non-ideal wavelength converters. Limited wavelength conversion has been shown to provide performance which is often close to that achieved with ideal wavelength conversion in networks with tunable transmitters and receivers.

Experience in measuring internet backbone traffic variability: Models metrics, measurements and meaning

Understanding the variability of Internet traffic in backbo ne networks is essential to better plan and manage existing networks, as well as to design next generati on networks. However, most traffic analyses that might be used to approach this problem are based on detai led packet or flow level measurements, which are usually not available throughout a large network. As a result there is a poor understanding of backbone traffic variability, and its impact on network oper ations (e.g. on capacity planning or traffic engineering). This paper introduces a metric for measuring backbone traffic variability that is grounded on simple but powerful traffic theory. What sets this metric a part, however, is that we present a method for practical implementation of the metric using widely availa ble SNMP traffic measurements. In addition to simulations, we use a large set of SNMP data from an operati onal IP network on the order of 1000 nodes to test our methods. We also delve into the degree and so urce of variability in real backbone traffic, providing insight into the true nature of traffic var i bility.

Load balancing and control for distributed World Wide Web servers

We describe an admission control and load balancing scheme for multicomputer World Wide Web servers. The scheme consists of a frontend admission control function which regulates requests sent to backend Web server using periodic load measurements passed from the backends to the frontends. The aims of scheme are two-fold: to ensure that traffic offered to backends can be serviced promptly, and to make full use of the backend capacity. We describe simulation results and an Erlang based implementation of the scheme. It is shown that the algorithm is trivial to implement, has negligible computational complexity, and provides effective load balancing and control. The scheme has now been implemented in the Eddie open source project, an Ericsson sponsored effort aimed at delivering commercial grade, quality of service driven Web server solutions.

Computing queue-length distributions for power-law queues

The interest sparked by observations of long-range dependent traffic in real networks has lead to a revival of interest in non-standard queueing systems. One such queueing system is the M/G/1 queue where the service-time distribution has infinite variance. The known results for such systems are asymptotic in nature, typically providing the asymptotic form for the tail of the workload distribution, simulation being required to learn about the rest of the distribution. Simulation however performs very poorly for such systems due to the large impact of rare events. We provide a method for numerically evaluating the entire distribution for the number of customers in the M/G/1 queue with power-law tail service-time. The method is computationally efficient and shown to be accurate through careful simulations. It can be directly extended to other queueing systems and more generally to many problems where the inversion of probability generating functions complicated by power-laws is at issue. Through the use of examples we study the limitations of simulation and show that information on the tail of the queue-length distribution is not always sufficient to answer significant performance questions. We also derive the asymptotic form of the number of customers in the system in the case of a service-time distribution with a regularly varying tail (e.g. infinite variance) and thus illustrate the techniques required to apply the method in other contexts.

Performance of networks using wavelength converters based on four-wave mixing in semiconductor optical amplifiers

In this paper, we examine the blocking performance of networks in which connections may be blocked due to either insufficient capacity or due to limitations in the transmission network. We use analytical expressions and network simulations to examine blocking in networks in which the quality of the received signal may be so poor that the connection is effectively blocked. In particular, we apply our analysis to networks which use wavelength converters based on four-wave mixing (FWM) in semiconductor optical amplifiers. We show that the performance improvements obtained using these wavelength converters can be significant, but this depends on whether the network uses fixed-frequency or tunable transmitters and receivers.

Impact of flow control on quality of service driven packet scheduling disciplines

In this paper we investigate the interaction between window based flow control and a recently proposed packet scheduling discipline designed for real-time services. The scheduling discipline, called the dual queue discipline, has been shown to provide greater flexibility than other scheduling disciplines such as fair queueing. However, its performance on non real-time services has not been previously investigated. We show that the dual queues performance for non real-time services is again better than that of alternative approaches, which indicates that it will perform well in an environment of mixed real-time and non real-time traffic, such as the current Internet.

Wavelength conversion in networks with differing link capacities

Transmission limitations, asymmetric loads and network upgrade strategies are all expected to mean that different links in a wavelength division multiplexed (WDM) network are expected to have different numbers of wavelengths. We examine the performance improvements offered by wavelength converters in WDM networks which have different number of wavelengths on different links. We show that wavelength converters can provide more significant performance improvements in networks with differing numbers of wavelengths on each link than in networks with the same number of wavelengths on each link. However, these performance improvements depend on the number of wavelengths on each link, the offered loads and the wavelength assignment scheme used.

Critical Congestion Control Issues in the Evolution of Common Channel Signaling Networks

Abstract The widespread deployment of Common Channel Signaling Networks (CCSNs) using Signaling System Number 7 (SS7) poses many challenges to telephone service providers. To support the planning process and ensure the reliability and robustness of the network, it is critical to understand the performance of the network both at engineered traffic levels, and during periods of failure and congestion. This paper discusses a number of critical traffic issues in the evolution SS7 CCSNs. In some cases these have received no attention in a CCSN context in the public literature. We shall show, through examples, that these factors play a major role in the grade-of-service provided by a network and that there still exist vast challenges for performance analysts in this area. In particular, an answer is required to a seemingly innocuous question: “WHAT ARE THE OBJECTIVES OF OVERLOAD CONTROL IN A CCS NETWORK?”.

An investigation of CCS tandem overload control issues

In this paper we analyze issues associated with overloads in tandem switching systems in common channel signaling (CCS) networks. In particular, we examine the need for, and effectiveness of, automatic congestion control (ACC), a CCS mechanism which allows switching systems to inform originating exchanges of overloaded processing resources and thereby control call rates to congested nodes. We demonstrate the critical need for end-offices and tandem switching systems to implement some form of ACC if network performance is to be maintained during congestion in tandem exchanges. In addition, we find that the presently defined ACC should be expanded to allow more levels of congestion to be reported to originating nodes, so that a more effective, finer grained control is produced.

A comparison of CCS tandem overload controls

In this paper we analyse issues associated with overloads in tandem switching systems in Common Channel Signalling (CCS) networks. In particular, we examine the need for, and effectiveness of, Automatic Congestion Control (ACC), a CCS mechanism which allows a congested switching system to inform adjacent CCS nodes of its level of overload, and subsequently controls the rate at which traffic is sent to the congested node. We demonstrate the critical need for end-offices and tandem switching systems to implement some form of ACC if network performance is to be maintained during congestion, especially in tandem exchanges. In addition, we find that the presently defined ACC should be expanded to allow more levels of congestion to be reported to adjacent nodes, so that a more effective, finer grained control is produced.