collaborates with the research team at the DAIC in INFOTEC.Abstract-The performance of a distributed file system is defined by its hardware components, as well as its operational parameters. Even a slight change on a working condition may induce the major impact, for instance, on service response times. In this paper, we propose a set of experiments on the Babel file system using a client that sends requests for either file storage or retrieval, with two file sizes (512 MB and 1 GB). The aim is to achieve the best working conditions. To measure the performance of the Babel system we took the throughput and the response time when any of two operations (storage or retrieval) was running. The analysis of results showed that, for a given set of operational parameters, there is an optimal file size that gets the best out of the system's performance.keywords-redundancy, storage, Babel file system, availability.

The Babel File System is a massive storage system, which is made up from three main components: clients, proxy and storage nodes. Each client perceives a single computer, called proxy, which dispatches all service requests randomly to an arbitrary storage node and manages the metadata. However, availability can be compromised if the proxy suddenly halts. To avoid this problem we propose the utilization of a redundant set of proxies. Nevertheless, this solution implies dealing with a consensus problem in order to guarantee the consistency of the metadata copies recorded at each and every proxy. In this paper we propose the use of the Paxos algorithm in order to overcome this problem. Our proposal can be understood as a middleware through which the proxies communicate among themselves to enforce metadata consistency. The experiments showed that, under certain random conditions, the progress of the protocol could be threatened. We propose the mechanisms to address these issues to set the system back to its stable state. However, stability is strongly limited by the correlation between failure and recovery rates.

Inductive logic programming (ILP) induces concepts from a set of positive examples, a set of negative examples, and background knowledge. ILP has been applied on tasks such as natural language processing, finite element mesh design, network mining, robotics, and drug discovery. These data sets usually contain numerical and multivalued categorical attributes; however, only a few relational learning systems are capable of handling them in an efficient way. In this paper, we present an evolutionary approach, called Grouping and Discretization for Enriching the Background Knowledge (GDEBaK), to deal with numerical and multivalued categorical attributes in ILP. This method uses evolutionary operators to create and test numerical splits and subsets of categorical values in accordance with a fitness function. The best subintervals and subsets are added to the background knowledge before constructing candidate hypotheses. We implemented GDEBaK embedded in Aleph and compared it to lazy discretization in Aleph and discretization in Top‐down Induction of Logical Decision Trees (TILDE) systems. The results obtained showed that our method improves accuracy and reduces the number of rules in most cases. Finally, we discuss these results and possible lines for future work.