Abstract
This paper presents our experience in coordinating and teaching a novel graduate systems and computing course named 'Successful Systems in Production' (SSP). The course targets graduate students of different research interests in Computer Science. The course aims at giving a breadth knowledge on cutting-edge well-known systems in production, and exploring the potential synergies across different areas of research. Having its roots in Distributed Computing, SSP addresses those systems that overlap with other research areas like Computational Systems, Parallel Computing, Databases, Cloud Computing, Artificial Intelligence, Security, etc. SSP exhibits an agile topic selection model that fits several students' backgrounds in each academic year. The topics focus on the practical aspects of each selected system that is considered 'successful', i.e., based on its worldwide impact and technical significance. This is important for graduate students to acquire best practices in industry and academia, necessary to build practical computing systems. In the same vein, the assessment method includes a project that is based on one of the presented systems and also intersects with the student's own research plan. Based on our teaching experience and the excellent feedback of the students, we strongly recommend this graduate course to be taught at other universities. © 2019 IEEE.

Abstract
Several millions of execution flows will be executed in ultrascale computing systems (UCS), and the task for the programmer to understand their coherency and for the runtime to coordinate them is unfathomable. Moreover, related to UCS large scale and their impact on reliability, the current static point of view is not more sufficient. A runtime cannot consider to restart an application because of the failure of a single node as statically several nodes will fail every day. Classical management of these failures by the programmers using checkpoint restart is also too limited due to the overhead at such a scale. The article explores programming models and runtimes required to facilitate the task of scaling and extracting performance on continuously evolving platforms, while providing resilience and fault-tolerant mechanisms to tackle the increasing probability of failures throughout the whole software stack. © The Institution of Engineering and Technology 2019.

Abstract

Abstract
Fog/Edge computing improves the latency and security of data by keeping storage and computation close to the data source. Nevertheless, this raises other security challenges against malicious, a.k.a, Byzantine, attacks that can exploit the isolation of nodes, or when access to distributed data is required in untrusted environments. In this work, we study the feasibility of deploying Byzantine Agreement protocols to improve the security of fog/edge systems in untrusted environments. In particular, we explore existing Byzantine Agreement protocols, heavily developed in the Blockchain area, emphasizing the Consistency, Availability, and Partition-Tolerance tradeoffs in a geo-replicated system. Our work identifies and discusses three different approaches that follow the Strong Consistency, Eventual Consistency, and Strong Eventual Consistency models. Our conclusions show that Byzantine Agreement protocols are still immature to be used by fog/edge computing in untrusted environment due to their high finality latency; however, they are promising candidates that encourage further research in this direction. © 2021, Springer Nature Switzerland AG.

Abstract

Abstract
Notable Byzantine Fault Tolerant protocols have been designed so far. These protocols are often evaluated on simple benchmarks, and few times on NFS systems. On the contrary, studies that addressed the behavior of BFT on large back-ends, like Directories, are few. We believe that studying such systems is crucial for practice community due to their popularity. In this paper, we integrate BFT with OpenLDAP Directory. We introduce the design of the integrated system, that we call BFT-LDAP. Then, we study its behavior accompanied with some useful observations. In addition, we discuss the cost overhead of this integration. Our approach ensures that OpenLDAP legacy code remains completely intact, and that the integration with BFT is straightforward using APIs. Moreover, we convey that the additional performance cost of BFT-LDAP is negligible as compared to that of stand-alone OpenLDAP. We conducted our experiments on Emulab. The experiments indicate that the performance discrepancy of BFT-LDAP is negligible whenever different state-of-the-art BFT protocols are used. Other experiments demonstrate that a little sacrifice in throughput (less than 10%) is needed in order to leverage the resiliency of OpenLDAP against Byzantine faults (i.e., through applying BFT). © 2012 IEEE.