Alessio Guerrieri

Distributed estimation of global parameters in delay-tolerant networks

Distributed estimation of global parameters in intermittently connected mobile environments is a challenging problem. In this paper, we introduce a set of methods, based on gossip techniques and population protocols, for performing such task. The applicability of such techniques to various environments, characterized by different mobility patterns, is evaluated through numerical simulations and discussed extensively. Guidelines are provided to help practitioners choosing the right method for their specific application problem.

DFEP: Distributed Funding-Based Edge Partitioning

As graphs become bigger, the need to efficiently partition them becomes more pressing. Most graph partitioning algorithms subdivide the vertex set into partitions of similar size, trying to keep the number of cut edges as small as possible. An alternative approach divides the edge set, with the goal of obtaining more balanced partitions in presence of high-degree nodes, such as hubs in real world networks, that can be split between distinct partitions. We introduce dfep, a distributed edge partitioning algorithm based on the metaphor of currency distribution. Each partition starts from a random edge and expands independently by spending currency to buy neighboring edges. After each iteration, smaller partitions receive an higher amount of currency to help them recover lost ground and reach a similar size to the other partitions. Simulation experiments show that dfep is efficient and obtains consistently balanced partitions. Implementations on both Hadoop and Spark show the scalability of our approach.

Top-k Item Identification on Dynamic and Distributed Datasets

The problem of identifying the most frequent items across multiple datasets has received considerable attention over the last few years. When storage is a scarce resource, the topic is already a challenge; yet, its complexity may be further exacerbated not only by the many independent data sources, but also by the dynamism of the data, i.e., the fact that new items may appear and old ones disappear at any time. In this work, we provide a novel approach to the problem by using an existing gossip-based algorithm for identifying the k most frequent items over a distributed collection of datasets, in ways that deal with the dynamic nature of the data. The algorithm has been thoroughly analyzed through trace-based simulations and compared to state-of-the-art decentralized solutions, showing better precision at reduced communication overhead.

DynamicDFEP: A Distributed Edge Partitioning Approach for Large Dynamic Graphs

Distributed graph processing has become a very popular research topic recently, particularly in domains such as the analysis of social networks, web graphs and spatial networks. In this context, graph partitioning is an important task. Several partitioning algorithms have been proposed, such as DFEP, JABEJA and POWERGRAPH, but they are limited to static graphs only. In fact, they do not consider dynamic graphs in which vertices and edges are added and/or removed. In this paper, we propose a graph partitioning method for large dynamic graphs. We present an implementation of the proposed approach on top of the AKKA framework, and we experimentally show that our approach is efficient in the case of large dynamic graphs.

Making puzzles green and useful for adaptive identity management in large-scale distributed systems

Various online systems offer a lightweight process for creating accounts (e.g., confirming an e-mail address), so that users can easily join them. With minimum effort, however, an attacker can subvert this process, obtain a multitude of fake accounts, and use them for malicious purposes. Puzzle-based solutions have been proposed to limit the spread of fake accounts, by establishing a price (in terms of computing resources) per identity requested. Although effective, they do not distinguish between requests coming from presumably legitimate users and potential attackers, and also lead to a significant waste of energy and computing power. In this paper, we build on adaptive puzzles and complement them with waiting time to introduce a green design for lightweight, long-term identity management; it balances the complexity of assigned puzzles based on the reputation of the origin (source) of identity requests, and reduces energy consumption caused by puzzle-solving. We also take advantage of lessons learned from massive distributed computing to come up with a design that makes puzzle-processing useful. Based on a set of experiments, we show that our solution provides significant energy savings and makes puzzle-solving a useful task, while not compromising effectiveness in limiting the spread of fake accounts.

Tovel: Distributed Graph Clustering for Word Sense Disambiguation

Word sense disambiguation is a fundamental problem in natural language processing (NLP). In this problem, a large corpus of documents contains mentions to well-known (non-ambiguous) words, together with mentions to ambiguous ones. The goal is to compute a clustering of the corpus, such that documents that refer to the same meaning appear in the same cluster, subsequentially, each cluster is assigned to a different semantic meaning. In this paper, we propose a mechanism for word sense disambiguation based on distributed graph clustering that is incremental in nature and can scale to big data. A novel, heuristic vertex-centric algorithm based on the metaphor of the water cycle is used to cluster the graph. Our approach is evaluated on real datasets in both centralized and decentralized environments.