Pedro Rodríguez

Testing a Cloud Provider Network for Hybrid P2P and Cloud Streaming Architectures

The number of online real-time streaming services deployed over network topologies like P2P or centralized ones has remarkably increased in the recent years. This has revealed the lack of networks that are well prepared to respond to this kind of traffic. A hybrid distribution network can be an efficient solution for real-time streaming services. This paper contains the experimental results of streaming distribution in a hybrid architecture that consist of mixed connections among P2P and Cloud nodes that can interoperate together. We have chosen to represent the P2P nodes as Planet Lab machines over the world and the cloud nodes using a Cloud providers network. First we present an experimental validation of the Cloud infrastructures ability to distribute streaming sessions with respect to some key streaming QoS parameters: jitter, throughput and packet losses. Next we show the results obtained from different test scenarios, when a hybrid distribution network is used. The scenarios measure the improvement of the multimedia QoS parameters, when nodes in the streaming distribution network (located in different continents) are gradually moved into the Cloud provider infrastructure. The overall conclusion is that the QoS of a streaming service can be efficiently improved, unlike in traditional P2P systems and CDN, by deploying a hybrid streaming architecture. This enhancement can be obtained by strategic placing of certain distribution network nodes into the Cloud provider infrastructure, taking advantage of the reduced packet loss and low latency that exists among its datacenters.

Predicting Stock Market Indices Movements

This paper examines the extent to which the daily movements of three large emerging markets stock indices are predictable. Lagged technical indicators are used as explanatory variables. In the analysis we employed seven classification techniques and assessed the discriminatory power of the classifiers through the area under the receiver operating characteristic (ROC) curve. The results show that the daily movements of the three indices are better predictable than random. After taking into account the bias induced by non-synchronous price quotations, a trading system with break-even costs is simulated. The non-random classifiers yield returns above those of both random walk and contrarian investment strategies. No inefficiency is found due to the fact that relatively low break-even transaction costs are enough to eliminate the sources of trading profits.

Materialising a new architecture for a distributed MCU in the Cloud

New technologies are making videoconferencing more ubiquitous than ever. This imposes a big challenge for scaling software MCUs, the traditional videoconferencing servers. We propose, implement and test an architecture for a distributed MCU designed to be deployed in a Cloud Computing environment. The main design idea is to break monolithic MCUs into more simple parts: broadcasters. These broadcasters can be deployed independently on the fly. This achieves a higher deployment granularity and flexibility. We describe the control architecture that allows this distribution and prove the viability of the system with a fully developed implementation. We propose a distributed architecture for video conferencing servers (MCUs).The MCU is divided into simple parts that broadcast streams, OneToManys.We describe and evaluate the new control architecture that allows distributed deployment.This solution provides improved scalability and deployment granularity.

dOTM: A Mechanism for Distributing Centralized Multi-party Video Conferencing in the Cloud

One of the key factors for a given application to take advantage of cloud computing is the ability to scale in an efficient, fast and reliable way. In centralized multi-party video conferencing, dynamically scaling a running conversation is a complex problem. In this paper we propose a methodology to divide the Multipoint Control Unit (the video conferencing server) into more simple units, broadcasters. Each broadcaster receives the media from a participant, processes it and forwards it to the rest. These broadcasters can be distributed among a group of CPUs. By using this methodology, video conferencing systems can scale in a more granular way, improving the deployment.

Linkages in international stock markets: evidence from a classification procedure

In this article, we propose a new approach to evaluate the predictable components in stock indices using a boosting-based classification technique, and we use this method to examine causality among the three main stock market indices in the world during periods of large positive and negative price changes. The empirical evidence seems to indicate that the Standard & Poors 500 index contains incremental information that is not present in either the FTSE 100 index (Financial Times Stock Exchange Index) or the Nikkei 225 index, and that could be used to enhance the predictability of the large positive and negative returns in the three main stock market indices in the world. This in turn would suggest a causality relationship running from the Standard & Poors 500 index to both the FTSE 100 and the Nikkei 225 indices.

Adaptive cross-device videoconferencing solution for wireless networks based on QoS monitoring

The increase in CPU power and screen quality of todays smartphones as well as the availability of high bandwidth wireless networks has enabled high quality mobile videoconferencing never seen before. However, adapting to the variety of devices and network conditions that come as a result is still not a trivial issue. In this paper, we present a multiple participant videoconferencing service that adapts to different kind of devices and access networks while providing an stable communication. By combining network quality detection and the use of a multipoint control unit for video mixing and transcoding, desktop, tablet and mobile clients can participate seamlessly. We also describe the cost in terms of bandwidth and CPU usage of this approach in a variety of scenarios.

Videoconference Capacity Leasing on Hybrid Clouds

This paper proposes a new methodology focused on implementing cost effective architectures on Cloud Computing systems. With this methodology the paper presents some disadvantages of systems that are based on single Cloud architectures and gives some advices for taking into account in the development of hybrid systems. The work also includes a validation of these ideas implemented in a complete videoconference service developed with our research group. This service allows a great number of users per conference, multiple simultaneous conferences, different client software (requiring transcodification of audio and video flows) and provides a service like automatic recording. Furthermore it offers different kinds of connectivity including SIP clients and a client based on Web 2.0. The ideas proposed in this article are intended to be a useful resource for any researcher or developer who wants to implement cost effective systems on several Clouds.

Using Machine Learning Algorithms to Find Patterns in Stock Prices

We use a machine learning algorithm called Adaboost to find direction-of-change patterns for the S&P 500 index using daily prices from 1962 to 2004. The patterns are able to identify periods to take long and short positions in the index. This result, however, can largely be explained by first-order serial correlation in stock index returns.

A Methodology for Designing and Evaluating Cloud Scheduling Strategies in Distributed Videoconferencing Systems

Over the last few years, videoconferencing systems have experienced several changes that enable videoconferencing applications in personal devices to a high number of users. To efficiently attend this high and variable demand, deploying distributed videoconferencing servers in cloud-based infrastructures is highly recommended. However, videoconferencing systems have particular characteristics that impede the application of regular resource scheduling solutions used in other kinds of distributed environments. In this paper, we propose a methodology to design and evaluate scheduling strategies adapted to the necessities of each specific scenario. It involves using a new metric to estimate the resource consumption of each connection and provides a set of coefficients to evaluate the efficiency of the strategy. We test the methodology in a real setup and compare the behavior and performance of three scheduling algorithms. The conclusion is that the proposed methodology allows us to configure decision policies adapted to the requirements and necessities of a range of use cases. Thereby, we achieve a more efficient way of using cloud resources, improving the service performance and saving costs.

A Metric to Estimate Resource Use in Cloud-Based Videoconferencing Distributed Systems

Scheduling resources in Cloud distributed videoconferencing systems presents a complex challenge not resolved yet. Traditional scheduling models are not applicable due to the particular characteristics of such type of systems. One of the main issues is estimating how many resources will consume a new client that connects to a videoconferencing session. Otherwise, it is difficult to decide where to allocate new requests. This paper proposes a new metric to perform this estimation basing on different parameters of the sessions. To validate the metric we set up a real scenario comparing the behaviour with and without the proposed metric. The conclusion is that the metric enables the design of more advanced and precise scheduling algorithms. Furthermore, thanks to this metric, resources are used more efficiently resulting in performance improvements and cost saving.