1st Workshop on New Challenges of Distributed Systems

Invited speakers

Sanjay Chawla (University of Sydney) Proximal Methods for Distributed Machine Learning

We will present several concrete examples of how proximal methods and geometric thinking can be used to address distributed machine learning problems in a principled fashion. Examples will include classification, non-negative matrix factorisation and dimensionality reduction.

Alan Fekete (University of Sydney) Consistency Properties of Cloud Data Stores

There is a very diverse set of cloud-deployed systems that provide reliable storage for huge amounts of data; some are offered as services by IaaS providers, and others are run as application software by consumers within the cloud environment. Examples include SimpleDB, Cassandra, MongoDB, Azure SQL database. These systems vary in the extent to which data consistency is supported; many aim at some form of eventual consistency. We report on a number of research results aimed at understanding the consistency properties and their implications for applications that run on top of these stores. We describe experiemental investigation of the properties provided by different systems, novel definitions that capture known properties, and proposals for new properties that might be supported in the future.

Rachid Guerraoui (EPFL) Eventual Consistency: Does it Make Sense?

The aim of this talk is to discuss several misunderstanding underlying the notion of eventual consistency.

Dali Kaafar (NICTA) Privacy in Content Oriented Networking: New Challenges, Threats and Countermeasures

As the Internet struggles to cope with scalability, mobility, and security issues, new network architectures are being proposed to better accommodate the needs of modern systems and applications. In particular, Content-Oriented Networking (CON) has emerged as a promising next-generation Internet architecture: it sets to decouple content from hosts, at the network layer, by naming data rather than hosts. CON comes with a potential for a wide range of benefits, including reduced congestion and improved delivery speed by means of content caching, simpler configuration of network devices, and security at the data level. However, it remains an interesting open question whether or not, and to what extent, this emerging networking paradigm bears new privacy challenges. In this talk, we aim to provide a systematic privacy analysis of CON and the common building blocks among its various architectural instances in order to highlight emerging privacy threats, and analyse a few potential countermeasures. Finally, we present a comparison between CON and s Internet in the context of a few privacy concepts, such as, anonymity, censoring, traceability, and confidentiality. In essence, this talk focuses on the identification of the new privacy challenges set by the adoption of next generation Internet architectures.

Anne-Marie Kermarrec (INRIA) Privacy-aware micro-blogging

The increasing use of micro-blogging sites have raised concerns over the arrests of the users involved, both in, generation and spread of information on several social networking and micro-blogging platforms. We will present an approach ensuring users to freely contribute in the spread of a particular piece of information while at the same time not affecting the generation of information on such platforms, typically making it impossible for an attacker to identify whether a user forwarded a particular information with certainty.

Zahir Tari (RMIT) Task Assignment in Distributed Systems - beyond traditional approaches

There has been an important body of knowledge in the area of performance in distributed systems. However most of the existing approaches do not take into account the specific nature of the underlying traffic (e.g. high variation in task size). Such characteristics are typically found in Web and Cloud traffic. Existing approaches "keep" using the exponential distribution for task size for a purpose of making task assignment much more simpler to design/implement. Obviously such assumptions are not realistic, and much more work is needed to come up with new ways to improve performance based on realistic assumptions of the traffic (e.g. heavy-tail distribution). In this talk I will provide an overview of some of the key approaches for load balancing in distributed systems, and later show some of the new ways to deal with busty traffic.

Albert Zomaya (University of Sydney) Adaptable Resource Allocation in Cloud Computing Systems

Cloud Computing is among the fastest growing topics in computing research today. Although the idea of offering computational power as a service is not novel, it did not publically become available before Clouds. Clouds are fundamentally different from their predecessors (Grids/Clusters) and thus require specific expertise to be properly utilized. Today, there are literally hundreds of cloud providers in all forms and scales. Computing needs (applications) have also become increasingly diverse. In this talk, I will describe challenges with the dynamicity and heterogeneity of resources and the diversity of applications from resource management perspective. Then, I will discuss how we can turn these challenges into opportunities for cloud systems' efficiency.