Accepted Doctoral Symposium papers

When: Tuesday December 13
Where: Department of Sociology and Social Research (via Verdi, 26) - Professor Room (Sala Professori)

The following are the papers accepted in the Doctoral Symposium. The abstracts can be visualized by clicking on the paper title.

• Algebraic Transformation of Descriptive Vector Byte-code Sequences

  Mads Ohm Larsen (Niels Bohr InstituteNiels Bohr Institute)

  Both high-productivity and high-performance are two often sought after aspects of scientific programming. Python gives the programmer high-productivity, but even with NumPy it is often not high-performant because of the GIL, which makes it inherently single threaded. Bohrium grants Python the ability to be easily run on multicore systems or GPUs. Since Bohrium is non-invasive, the programmer still gets the same high-productivity from Python that he is used to. Further investigation into optimizing the Bohrium byte-code is needed. One approach is using algebraic transforms on the byte-code sequences.

• Self-stabilizing Middleware Services

  Ioannis Marcoullis (University of Cyprus)

  We present a set of middleware services with the fault-tolerant property of self-stabilization. Services with this property can return the system to its desired behavior in the case where the system’s state is corrupted by transient faults. We first present a self-stabilizing virtual synchrony service realize state machine replication (SMR), in a setting with a fixed set of processors. We then develop a multi-purpose reconfiguration service that enables the above SMR module and other services to be deployed in settings where the processor set is dynamic. The current work provides the theoretical basis to move into prototyping and evaluation.

• AdaptCache: Adaptive Data Partitioning and Replication for Distributed Object Caches

  Omar Asad (McGill University)

  This paper presents the AdaptCache project. AdaptCache is an adaptive caching middleware for application servers that monitors the current workload and generates policies to distribute and/or replicate objects and requests among the local caches of application servers so that most requests can be execute on locally cached objects and, at the same time, the load will be evenly distributed among servers. The project is divided into two main phases. The first one, which is described in detail in this paper, tackles the problem of dynamically distributing objects and requests for volatile and fluctuating e-commerce applications. Several data distribution approaches based on graph partitioning are proposed. The approaches are compared using the YCSB and RUBiS benchmarks showing that AdaptCache is able to dynamically capture various workload changes and react quickly to these changes. The second phase of the AdaptCache project explores data replication for distributed object caches. It discusses the advantages of object replication such as increased locality but also possible overheads due to consistency requirements and space limitations. Any dynamic replication solution must take these issues into account.

• Privacy-enhancing Federated Middleware for the Internet of Things

  Paul Fremantle (University of Portsmouth)

  The Internet of Things (IoT) offers new opportunities, but alongside those come many challenges for security and privacy. Most IoT devices offer no choice to users of where data is published, which data is made available and what identities are used for both devices and users. The aim of this work is to explore new middleware models and techniques that can provide users with more choice as well as enhance privacy and security. This paper outlines a new model and a prototype of a middleware system that implements this model.

• The Prospects for Multi-Cloud Deployment of SaaS Applications with Container Orchestration Platforms

  Vincent Reniers (KU Leuven)

  Recent years have seen an increased adoption of container technology for software deployment and lightweight virtualization. More recently, container orchestration systems provide a platform for container deployment and management of cluster resources. Software-as-a-Service (SaaS) providers traditionally make use of middleware to facilitate multi-tenancy in a federated cloud. Container orchestration presents many opportunities in achieving scalability and providing cost-efficient multi-tenancy. In this paper, we outline opportunities and challenges for multi-cloud deployment of containerized SaaS applications.

• Service Platform for Continuous Delivery of Assisted Living Systems

  João Bentes (School of Information Technology, Halmstad University)

  A smart home has potential to support independent living of elderly people in their preferred living environments. However, smart home systems do not fully address the aims of Ambient assisted living (AAL), mainly due to limited support outside the home. This need of continuous delivery of assistance for elderly people on the go require technology which extends the home into the society. This ongoing work proposes to identify the architectural requirements for a service platform being able to continuously deliver assistive services at home and beyond.

• Distributed and cooperative intrusion detection in cloud networks

  Saadia Ghribi (ENISENIS)

  Cloud Computing is remained the most potential technology. It offers numerous benefits to both end users and businesses. However, it exposes many security challenges and vulnerabilities because of multi-tenancy and virtualization. Thus, security is becoming the biggest concern that can influence the adoption of Cloud technology. Hence, Intrusion detection is becoming a necessity in order to secure cloud environment. However, the commonly used IDS bring some issues that the most common one is the absence of cooperative layer detection. Our approach consists in proposing a new middleware IDS based on cooperation between cloud computing layers.

• Pyramid Codes to the rescue: A preliminary evaluation for cloud storage systems

  Roberta Barbi (Université de Neuchâtel)

  In this paper, we study the tradeoffs between reliability, storage overhead, repair locality, and performance of erasure-correcting codes implemented in real-world distributed storage systems. In particular, we focus on a recent family of codes with locality called “pyramid codes”. We first describe the properties of pyramid codes and their possible applications. We then demonstrate, based on a preliminary evaluation of a Python implementation, the benefits of the good locality properties of the pyramid codes, which we plan to leverage in a distributed storage system.

• High Throughput Tertiary Storage in HPC Environments

  Klaus B. Jensen (Niels Bohr InstituteNiels Bohr Institute)

  Magnetic tape in High Performance Computing (HPC) environments has traditionally been used as a tertiary layer in a Hierarchical Storage Management (HSM) system. We propose that tape be given a more central role in the HPC data centre with direct involvement of the batch scheduler. To this end, we have begun development of a purpose-built tape storage system called Tapr. The write strategy used by Tapr allows the system to provide precise estimates on transfer time of data from tertiary storage to primary fast parallel storage.

• Position paper: Toward an holistic approach of Systems of Systems

  Simon Bouget (IRISA - Université de Rennes)

  Large scale distributed systems have become ubiquitous, from on-line social networks to the Internet-of-things. To meet rising expectations (scalability, robustness, flexibility,...) these systems increasingly espouse complex distributed architectures, that are hard to design, deploy and maintain. To grasp this complexity, developers should be allowed to assemble large distributed systems from smaller parts using a seamless, high-level programming paradigm. We present such an assembly-based programming framework, enabling developers to easily define and realize complex distributed topologies as a construction of simpler blocks (e.g. rings, grids). It does so by harnessing the power of self-organizing overlays, that is made accessible to developers through a high-level Domain Specific Language and self-stabilizing runtime. Our evaluation further shows that our approach is generic, expressive, low-overhead and robust.