Christian Nyberg

Web server performance modeling using an M/G/1/K*PS queue

Performance modeling is an important topic in capacity planning and overload control for web servers. We present an M/G/1/K*PS queueing model of a web server. The arrival process of HTTP requests is assumed to be Poissonian and the service discipline is processor sharing. The total number of requests that can be processed at one time was limited to K. Closed form expressions were obtained for web server performance metrics such as average response time, throughput and blocking probability. The average of the service time requirement and the limit of the number of requests being served were model parameters. The parameters were estimated by maximizing the log-likelihood function of the measured average response time. Compared to other models, this model is conceptually simple and it is easy to estimate model parameters. The model had been validated through lab measurements. The performance metrics predicted by the model fit well to the experimental outcome.

Exploring Bottlenecks in Market-Driven Requirements Management Processes with Discrete Event Simulation

This paper presents a study where a market-driven requirements management process is simulated. In market-driven software development, generic software packages are released to a market with many customers. New requirements are continuously issued, and the objective of the requirements management process is to elicit, manage, and prioritize the requirements. In the presented study, a specific requirements management process is modelled using discrete event simulation, and the parameters of the model are estimated based on interviews with people from the specific organisation where the process is used. Based on the results from simulations, conditions that result in an overload situation are identified. Simulations are also used to find process change proposals that can result in a non-overloaded process. The risk of overload can be avoided if the capacity of the requirements management process is increased, or if the number of incoming requirements is decreased, for example, through early rejection of low-priority requirements.

Performance simulation of a TINA network

The Telecommunications Information Networking Architecture (TINA) is based on object orientation and distributed computing. It is intended to allow a rapid and flexible development of new applications on all kinds of software and hardware platforms. In this paper we investigate the performance and capacity in a TINA network that supports one simple telephone service. Specifically, we examine how the network performance is affected by the distribution of computational objects among the physical nodes. We show that the distribution of objects has a significant effect on both the maximum allowed arrival rate and the mean set up time for a service. Also, we show that the heavy signalling needed to set up a service may result in performance problems in a TINA network.

Load Balancing for a Distributed CORBA-Based SCP

This paper examines load balancing issues relating to a distributed CORBA-based Service Control Point. Two types of load balancing strategies are explored through simulation studies: (i) A novel ant-based load balancing algorithm, which has been devised specifically for this type of system. This algorithm is compared to more traditional algorithms, (ii) A method for optimal distribution of the computational objects composing the service programs. This is based on mathematically minimising the expected communication flows between network nodes and message-level processing costs. The simulation model has been based on the recently adopted OMG IN/CORBA Interworking specification and the TINA Service Session computational object model.

Measuring Real-time Performance in Distributed Object Oriented Systems

The principles of distributed object oriented programming offer great possibilities for flexible architectures in multiple fields. In telecommunications, an architecture called Telecommunication Information Networking Architecture has been developed using these very principles. It allows telecommunication services to be implemented using software objects that in turn can be executed in a location transparent way in a network. The location transparency offers great flexibility for service creation, but as the software must be executed somewhere in the network on nodes of finite capacity, performance problems can arise due to inefficient placement of objects causing either overloaded nodes or excessive and unnecessary inter-node communication. To ensure good performance, various measures of load control and load balancing must be taken. We discuss how to measure the performance of a distributed object oriented system and examine two load balancing algorithms that can be used in such systems.

Modeling, implementation and evaluation of IEEE 802.11ac in NS-3 for enterprise networks

In this work we implement features for IEEE 802.11ac in the NS-3 simulator, in particular wider channels and bit-error calculations for higher modulation coding schemes. We also implement four wireless LAN deployment scenarios from the 802.11ax working group scenario document, and evaluate their performance under different operating conditions. Our simulation results demonstrate that many nodes in an enterprise network will yield lower average throughput to each AP and several APs on the same channel will create unreliable networks with some stations getting high throughput and some not able to send at all. Significant improvement in throughput was also observed with the use of frame aggregation.

A distributed MAC scheme to achieve QoS in ad hoc networks

Real-time applications introduce new requirements on wireless networks and impose quality thresholds on parameters like delay, jitter, throughput, and packet loss in order to run smoothly. This paper addresses this issue by presenting a MAC scheme that offers real-time applications the opportunity to reserve transmission time based on their QoS requirements for contention-free medium access. Our scheme, which is called EDCA with Resource Reservation (EDCA/RR), operates in a fully distributed manner, is compatible with IEEE 802.11, and provides both prioritized and parameterized QoS. In this study, we have extended EDCA/RR to handle reservation collisions and, through extensive simulations, we show that our proposal can handle multiple reservations as well as uninformed stations that lie outside the transmission range of both the transmitter and the receiver while providing QoS guarantees. We compare EDCA/RR with EDCA and our results show that, as the traffic in the network increases, EDCA/RR succeeds providing the required service to QoS-demanding applications whereas EDCA fails in this task. In addition, when the medium is lossy we show that, not only does EDCA/RR give better service to real-time traffic, but also to contending non-real-time traffic.

Design and Evaluation of an Overload Control System for Crisis-RelatedWeb Server Systems

During recent years we have seen several large-scale crises. The 9/11 terror attacks, tsunamis, storms, floods and bombings have all caused a great deal of damage. A common factor in these crises has been the need for information and one important source of information is usually Web sites. In this work we investigate and design an overload control system for Web sites that are vital in crises. The overload control system uses content adaption to dynamically control Web site performance

Reusable Simulation Models for Performance Analysis of Intelligent Networks

New services will be introduced at a much faster rate than today when Intelligent Networks are employed. It is important to control the process of introducing services and be able to forecast the consequences of a new service. This paper describes a method for performance simulation based on reusable simulation models which can be used for this purpose. The method builds on dividing the simulation model into three parts describing architecture, software of services and usage of the services respectively. Services and nodes can be described on different abstraction levels. If a new service is introduced all the old parts of the simulation program can be reused.

On overload control of high-speed switching nodes

In this report three basic principles for overload control of high speed telecommunication switches are presented, evaluated and compared. Then the concept of priorities is introduced into these models. Priorities are given to jobs related to calls being in progress. In a second case jobs associated to one out of many different services are prioritierised. The approach taken here is applicable to centralized processor systems as well as to a loosely coupled fully distributed processor environment.