Big Data has become a pervasive technology to manage the ever-increasing volumes of data. Among Big Data solutions, scalable data stores play an important role, especially, key-value data stores due to their large scalability (thousands of nodes). The typical workflow for Big Data applications include two phases. The first one is to load the data into the data store typically as part of an ETL (Extract-Transform-Load) process. The second one is the processing of the data itself. BigTable and HBase are the preferred key-value solutions based on range-partitioned data stores. However, the loading phase is inefficient and creates a single node bottleneck. In this paper, we identify and quantify this bottleneck and propose a tool for parallel massive data loading that solves satisfactorily the bottleneck enabling all the parallelism and throughput of the underlying key-value data store during the loading phase as well. The proposed solution has been implemented as a tool for parallel massive data loading over HBase, the key-value data store of the Hadoop ecosystem.

The modern healthcare landscape is overwhelmed by data derived from heterogeneous IoT data sources and Electronic Health Record (EHR) systems. Based on the advancements in data science and Machine Learning (ML), an improved ability to integrate and process the so-called primary and secondary data fosters the provision of real-time and personalized decisions. In that direction, an innovative mechanism for processing and integrating health-related data is introduced in this article. It describes the details of the mechanism and its internal subcomponents and workflows, together with the results from its utilization, validation, and evaluation in a real-world scenario. It also highlights the potential derived from the integration of primary and secondary data into Holistic Health Records (HHRs) and from the utilization of advanced ML-based and Semantic Web techniques to improve the quality, reliability, and interoperability of the examined data. The viability of this approach is evaluated through heterogeneous healthcare datasets pertaining to personalized risk identification and monitoring related to pancreatic cancer. The key outcomes and innovations of this mechanism are the introduction of the HHRs, which facilitate the capturing of all health determinants in a harmonized way, and a holistic data ingestion mechanism for advanced data processing and analysis.

Data streaming engines process data on the fly in contrast to databases that first, store the data and then, they process it. In order to process the increasing amount of data produced every day, data streaming engines run on top of a distributed system. In this setting failures will likely happen. Current distributed data streaming engines like Apache Flink provide fault tolerance. In this paper we evaluate the impact on performance of fault tolerance mechanisms of Flink during regular operation (when there are no failures) on a distributed system and the impact on performance when there are failures. We use the Intel HiBench for conducting the evaluation.

The need of systems able to deal with large amounts of data is increasing in the big data era. These systems run on top of a distributed system where failures will happen. In the last decade a large number of data stores have been developed for providing scalability. These systems avoid some of the properties traditional relational databases provided in order to achieve scalability producing a variety of data stores known as NoSQL data stores. In the last years, NewSQL systems have been developed in order to meet the best of both worlds: transactional processing and SQL language and scalability. In this paper we present a practical experience on how failures affect the regular operation of a scalable fault-tolerant relational database (NewSQL), CumuloNimbo. The paper evaluates the performance of CumuloNimbo using the TPC-C benchmark varying the number of nodes where the database runs under different failure scenarios.

Explainable AI (XAI) techniques are used for understanding the internals of the AI algorithms and how they produce a particular result. Several software packages are available implementing XAI techniques however, their use requires a deep knowledge of the AI algorithms and their output is not intuitive for non-experts. In this paper we present a framework, (XAI4PublicPolicy), that provides customizable and reusable dashboards for XAI ready to be used both for data scientists and general users with no code. The models, and data sets are selected dragging and dropping from repositories While dashboards are generated selecting the type of charts. The framework can work with structured data and images in different formats. This XAI framework was developed and is being used in the context of the AI4PublicPolicy European project for explaining the decisions made by machine learning models applied to the implementation of public policies.

Nowadays many applications are deployed in a distributed setting with components running on edge devices while others are running in (public/private) cloud or on-premises clusters to optimize performance and improve data privacy. The Computing Continuum is characterized by the seamless integration of diverse computing environments and devices. The integration of the required infrastructure to provide a seamless vision of all resources needs several software components that enable the vision of a continuum of the infrastructure in which applications are distributed and communicate with other components regardless of their location. The performance of applications in these scenarios will vary depending on how application components are distributed and which resources are provided. In this paper we describe the main components of a portable framework for the execution of container-based applications in the computing continuum. The framework is based on widely adopted open source components that can be deployed in private and public clouds, on-premises clusters and in resource-constrained devices. A preliminary performance evaluation in a deployment with on-premises and edge clusters is presented. The performance evaluation focuses on the latency perceived by clients according to client and service location, and service deployment time after application reconfiguration. The results of this evaluation will allow us to continue with the design and implementation of Vishnu, a platform for the computing continuum.