The performance of web caches, in Web Content Management Systems, can be improved by assembling only some of the content elements of a web page in the application server, and finishing the assembling process in the cache proxy. Due to this, the cache is able to manage parts of the web page instead of whole pages, which improves its performance. We propose an algorithm based on decision trees and obtained in a training process to create content fragmentation designs. Data mining is used in the training phase. Inputs of the classification algorithm must be monitored from the system producing small overheads. The paper contribution are the validation of: the use of classification system to self-adapt content fragmentation designs to improve the web performance, the parameters set to be used as inputs of the decision tree and finally, the suitability of using decision trees to represent and implement, in the classification system, the previous extracted knowledge. All these aspects are validated by experimental results extracted from a test-bed.

Performance model interchange formats provide a mechanism for automatically moving performance models among modelling tools. The Experiment Schema Extension (Ex-SE) specifies performance studies to be run on the model and the XML performance metrics desired. Moreover, the Results Schema Extension (Results-SE) specifies how to automatically transform the output metrics into useful results. This paper presents the tool FORGE (Friendly Output to Results Generator Engine), a GUI based transformation tool from output to results which simplifies the specification of common tables for the output files the user provides. A case study demonstrates the use of the tool.

XML-based interchange formats for performance models provide a mechanism whereby performance model information may be transferred among modeling tools. Formats have been defined for the interchange of queueing network models (QNM), layered queueing networks, UML, Petri nets (PNs) and others. These formats specify the model and a set of parameters for one run. For model studies, however, it is useful to be able to specify multiple runs, or experiments, for the model; to collect model output; and to present it in a useful, readable format for analysis and presentation. This paper presents a performance model interoperability framework that brings together performance model interchange formats and experiment specifications with the automatic generation of performance analysis results for presentation and publication. We present the output specifications, the requirements for the types of results to be supported, the issues in the output-to-results transformation, the results specifications and several prototype implementations that demonstrate the viability of the approach. We apply the approach to two distinct modeling paradigms: QNM and PN models. Several proof of concept experiments demonstrate the framework.

Ontology Matching (OM) consists of set of processes that decide and establish if ontology elements share lexical or semantic relationships. For a correct alignment, when we establish semantic relationships, it is necessary to find out the meaning of the elements that depends on the context. In this paper, we propose an extension of OWL called OWL-M to improve the representation of multiple types of relationships among ontology elements taking account the context. In function of each OWL-M alignment, a similarity degree is computed. Moreover, to assist in decision making when human participation is required, we define a set of annotations to describe the logical decisions used by OWL-M processes.

Companies and web sites store information about the user behaviour which is not shared with other people. A tool for public and free sharing of this information has been developed. This tool is installed as a Firefox Add-on and it stores, in a central system (Web Mining Service), the gathered information in the browser of the client.

We address two issues in this work: on one hand, we consider and compare five throughput predictors to be used in a Web system in order to estimate its future performance and, on the other hand, we propose an adaptive QoS-aware admission control algorithm that is based on a resource allocation scheme that includes a throughput predictor. In order to obtain a low overhead, the monitoring of the arrival traffic to the Web system is done following an adaptive time slot scheduling based on the burstiness factor that we defined in a previous work. Simulation results of the behaviour of the five throughput predictors and the admission control algorithm are presented and discussed.

Time-driven sensor networks are devoted to the continuous reporting of ambient data to the base station. In many cases, these data are provided by nodes that have been deployed in a structured manner, either by selecting strategic locations or by adopting some regular sampling pattern. In either case, the resulting inter-node distances may not be small, and thus additional supporting nodes may be necessary. This suggests that the problem could be better addressed from a network planning perspective. In this sense, a particular approach is proposed and, as part of it, the paper focuses on optimizing the lifetime that can be predicted from the network topology, which is assumed to be a static data gathering tree. It is shown that this problem requires the exploration of all possible spanning trees, since the energy consumed by a node depends on its workload, which in turn depends on how this node is connected to its neighborhood. Because this is an NP-hard problem, the use of a heuristic approach is required. Then, an algorithm based on simulated annealing is proposed, which converges asymptotically to the global optimum. This algorithm is tested on different scenarios and its computational complexity is proved to be linearly dependent on the number of nodes.

This paper presents a load balancing solution for a cluster of Web servers that considers QoS in order to differentiate the service provided to the clients when the Web system is receiving a high volume of traffic. Adaptability is a needed feature to suit the balancing parameters to the fluctuations of the Web traffic. We monitor the incoming workload to the system and tune the parameters more frequently when the bursty arrivals trend increases. Throughput prediction enables to estimate the maximum utilization allowed in the servers for each service, and then to set a maximum number of accepted requests to set a limit in the utilization of the servers and to avoid a possible congestion situation, while guaranteeing the QoS. Classical dispatching policies are introduced and compared to analyse the behaviour of the algorithm. HTTP static and dynamic requests are considered in the workload.

We present a Web system architecture using ontologies to improve the behavior of the system from the performance viewpoint. Since Web system performance indexes depend on state and parameter values on runtime period, the proposed system configuration will change during this period. In order to perform this change, the Web system is monitorized and gathered information stored into a knowledge base. We also model the performance of the different Web system elements intervening in the configuration using the knowledge base expressed by means of ontologies. An example of the use of this ontology in cache tier is also presented. We propose the use of performance reasoners to change the configuration during runtime period based on the information supplied from the knowledge base.

Several techniques and applications have been proposed to aid the decision taking process in the system performance domain. Most of these techniques have depicted the performance model of systems through annotations of performance measurements coming from specific software descriptive syntactical languages. However, the semantic representation of performance information provides the possibility of its ulterior machine-processable logical interpretation and therefore the applicability of inference rules about a particular domain. In this work, we propose a performance ontology together with the system performance analysis technique as an example of framework building for intelligent applications based on semantic Web

A Performance Model Interchange Format (PMIF) provides a mechanism whereby system model information may be transferred among queueing network model (QNM) based modeling tools. The PMIF allows diverse tools to exchange information and requires only that those tools provide importing/exporting mechanisms from/to the PMIF. The XML specification of the PMIF allows implementers to use widely available tools to parse the XML file, check the syntax, and simplify the translation to/from the XML format. Those tools, however, do not know the semantics of a QNM so they cannot check the XML to ensure that it contains a valid QNM. This paper presents the study of the validations needed to carry out such a semantic analysis, and the development of a semantic validation tool that can be used by any developer who wants to implement PMIF import/export mechanisms.

In this paper, an exact solution for the response time distribution of a single server, infinite capacity, discrete-time queue is presented. This queue is fed by a flexible discrete-time arrival process, which follows an on/off evolution. A workload variable is associated with each arrival instant, which may correspond to the service demand generated by a single arrival, or represent the number of simultaneous arrivals (bulk arrivals). Accordingly, the analysis focuses on two types of queues: (On/Off)/G/1 and (Batch-On/Off)/D/1. For both cases, a decomposition approach is carried out, which divides the problem into two contributions: the response time experienced by single bursts in isolation, and the increase on the response time caused by the unfinished work that propagates from burst to burst. Particularly, the solution for the unfinished work is derived from a Wiener-Hopf factorization of random walks, which was already used in the analysis of discrete GI/G/1 queues. Compared to other related works, the procedure proposed in this paper is exact, valid for any traffic intensity and has no constraints on the distributions of the input random variables characterizing the process: duration of on and off periods, and workload. From the general solution, an efficient and robust iterative algorithm for computing the expected response time of both queues is developed, which can provide results at any desired precision. This algorithm is numerically evaluated for different types of input distributions and proved against simulation.

Our research focuses on reducing the computing time in large simulation models of client-server architectures by replacing server models by load dependent servers (LDSs), whose solution requires less computational effort. As these LDSs include overhead information, it is necessary to design a set of measurements for collecting overhead data. We are proposing a methodology that indicates how to reduce the amount of measurements and how to model overhead by means of regression analysis. In this work we explain how the methodology is applied to a workload composed of TPC-C transactions that runs in a server with Solaris 2.5 operating system (OS). Also, we describe the construction of the LDS. We compare the LDS predictions with measurements of the TPC-C benchmark.

B-ISDN is conceived to support all types of existing and new emerging applications. In the context of B-ISDN and the new age of information, the user terminal will generate data, voice, video and, in general, multimedia traffic. ATM is the selected mechanism for the efficient transfer of these information across the future worldwide network. The proper design of this network requires cost-effective solutions that may satisfy the quality of service constraints of all the connections. To achieve this goal, an accurate characterization of the traffic generated by the sources is needed. This characterization should capture the statistical properties of the ATM traffic that are relevant to network performance. A review of literature shows that, from a general point of view, this objective has been reached, but with two drawbacks: traffic models are application dependant and, moreover, for some applications, traffic models are not analytically tractable. The final objective of this paper is to construct a general analytical B-ISDN model. Particularly, the paper focuses on the on-off models with general distributions, for which a complete exact statistical characterization at cell level was already obtained in a previous paper. On the basis of this characterization, an algorithm devoted to capture the statistical behaviour of any source is formulated and developed. Particularly, the algorithm captures the average rate and burstiness of any source in an exact way, and the autocorrelation function in an optimal way. Finally, numerical results for different types of video sources are provided.

The paper focuses on reducing completion time in large simulation models by replacing some of their elements with analytically determined solutions. In particular, we study realistic analytical server models that include operating system overhead. This overhead will be determined by measuring the behaviour of a representative workload. We propose a methodology to reduce the number of measurements based on information provided by an analytical workload model. The calibration of the model is also presented. As workload, the TPC-C benchmark was chosen due to our interest in transactional systems.

This paper presents a pool of performance models for user applications using IONA's OrbixTalk, a software system that provides a reliable multicast objects message system using UDP-based IP multicast. The solution presents models for all the levels involved, from the OrbixTalk user program level to the logical level regarding internal proxies using UDP sockets over TCP/IP ports. The OrbixTalk models have been included into the Models' Repository of HELIOS (Hierarchical Evaluation of Large Information Open Systems), a CASE tool for predicting the performance of distributed information systems.