Paula Prata

Algorithm based fault tolerance versus result-checking for matrix computations

Algorithm Based Fault Tolerance (ABFT) is the collective name of a set of techniques used to determine the correctness of some mathematical calculations. A less well known alternative is called Result Checking (RC) where, contrary to ABFT, results are checked without knowledge of the particular algorithm used to calculate them. In this paper a comparison is made between the two using some practical implementations of matrix computations. The criteria are performance and memory overhead: ease of use and error coverage. For the latter extensive error injection experiments were made. To the best of our knowledge, this is the first time that RC is validated by fault injection. We conclude that Result Checking has the important advantage of being independent of the underlying algorithm. It also has generally less performance overhead than ABFT, the two techniques being essentially equivalent in terms of error coverage.

Practical issues in the use of ABFT and a new failure model

We study the behavior of algorithm based fault tolerance (ABFT) techniques under faults injected according to a quite general fault model. Besides the problem of roundoff error in floating point arithmetic we identify two further weakpoints, namely lack of protection of data during input and output, and incorrect execution of the correctness checks. We propose the robust ABFT technique to handle those weakpoints. We then generalize it to programs that use assertions, where similar problems arise, leading to the technique of robust assertions, whose effectiveness is shown by fault injection experiments on a realistic control application. With this technique a system follows a new failure model, that we call fail-bounded, where with high probability all results produced are either correct or, if wrong, they are within a certain bound of the correct value, whose exact value depends on the output assertions used. We claim that this failure model is very useful to describe the behavior of many low redundancy systems.

The role of radio frequency identification (RFID) technologies in improving process management and product tracking in the textiles and fashion supply chain

Abstract: This chapter aims to gain a better understanding of RFID deployment in the Fashion and Textile Supply Chain (FTSC), mainly concerning its application in supporting operations, its advantages and main business drivers. In an attempt to find an RFID deployment pattern within the FTSC, a cross-case analysis was performed. The RFID experience of five companies based in different countries was analysed. From these empirical data, three main conclusions are drawn: (1) RFID is deployed mainly to support the following operations: the handling process; tracking work-in-progress; receiving operations; shipping operations; tracking products; tracking inventories; monitoring and sorting of merchandise; counting stock and picking merchandise; tracking containers; shipping; locating products; and store management. (2) The main advantages highlighted are: better inventory management; improved read rates; and order accuracy. (3) The main driver that led the case study companies to deploy RFID technology was the identification of inefficiencies in several of their operations.

OpenCL implementations of a genetic algorithm for feature selection in periocular biometric recognition

This paper explores OpenCL implementations of a genetic algorithm used to optimize the features vector in periocular biometric recognition. Using a multi core platform the algorithm is tested for CPU and GPU, exploring different parallelization levels for each operator of the genetic algorithm. The results show that using the GPU platform it is possible to accelerate the algorithm by several orders of magnitude, with a recognition rate similar to the one obtained in the sequential version. The results also show that it is possible to use only a small portion of the features without any degradation of the classifiers recognition rate.

Using the cloud to build a computing lab

The management of a laboratory for the many classes/areas in an education institution makes a set of challenges for the IT department which face a growing pressure to provide the necessary computing resources at the lowest price. This paper proposes a platform named CSCLab for the provisioning and management of laboratories with the computing resources necessary to teach a class using the cloud. The system was designed to be used by non-experts in cloud technologies and implemented to run in OpenStack but it can be easily adapted to other platforms with similar services.

A grid infrastructure for online games

Running Jake2 (Quake2 port to Java) on a grid is not a trivial task. For that purpose, we had to migrate a networking client-server model to a grid-based model. Online games are just computational systems that can benefit from this migration, in particular in the management and optimization of resources (i.e. processor and memory usage), as needed to get improved load balancing, more scalability, and low latency from the server side.

Management of virtual machine images in heterogeneous clouds

This paper presents the image server of the VISOR cloud agnostic virtual machine images management service. An evaluation approach is also described and the results are discussed. VISOR is not intended to fit in a specific cloud framework but rather to overreach sharing and interoperability limitations among different frameworks. Such feature is achieved by relying on a flexible image metadata schema and in a storage abstraction layer, where the latter seamlessly abstracts the heterogeneity of distinct cloud storage systems into which images can be saved. VISOR is an open source highly distributed system composed of several independent web services. In its development we have focused in modularity, extensibility and performance. Conducted evaluation tests have shown encouraging results regarding performance and scalability.

Towards cost-effective bio-inspired optimization: a prospective study on the GPU architecture

This paper studies the impact of varying the populations size and the problems dimensionality in a parallel implementation, for an NVIDIA GPU, of a canonical GA. The results show that there is an effective gain in the data parallel model provided by modern GPUs and enhanced by high level languages such as OpenCL. In the reported experiments it was possible to obtain a speedup higher than 140 thousand times for a populations size of 262 144 individuals.

Item Response Models in Computerized Adaptive Testing: A Simulation Study

In the digital world, any conceptual assessment framework faces two main challenges: (a) the complexity of knowledge, capacities and skills to be assessed; (b) the increasing usability of web-based assessments, which requires innovative approaches to the development, delivery and scoring of tests. Statistical methods play a central role in such framework. Item response models have been the most common statistical methods used to address such kind of measurement challenges, and they have been used in computer-based adaptive tests, which allow the item selection adaptively, from an item pool, according to the person ability during test administration. The test is tailored to each student. In this paper we conduct a simulation study based on the minimum error-variance criterion method varying the item exposure rate (0.1, 0.3, 0.5) and the test maximum length (18, 27, 36). The comparison is done by examining the absolute bias, the root mean square-error, and the correlation. Hypotheses tests are applied to compare the true and estimated distributions. The results suggest the considerable reduction of bias as the number of item administered increases, the occurrence of ceiling effect in very small size tests, the full agreement between true and empirical distributions for computerized tests of length smaller than the paper-and-pencil tests.

Database and State Replication in Multiplayer Online Games

Online games are nowadays part of an industry that involves several million dollars in transactions. Despite games are not critical applications in a traditional way, fault tolerance as a path to achieve high availability rates starts to emerge. A few hours or even days of game downtime can represent significant losses, either by subsequent player resignation or a strong credibility decrement. This article embodies a study on the cost of fault tolerance techniques, for both data and state replication, when applied to the well-known multiplayer game Dungeons & Dragons. The evaluation of the replication cost was done using the JMeter tool to simulate several simultaneous games. We reasoned out that although performance becomes smaller with the increase of connected players, that drawback is mainly related to the number of connections rather than replication itself.