Andrew Berns

Dissecting Self-* Properties

The scale and complexity of distributed systems have steadily grown in the recent years. Management of this complexity has drawn attention towards systems that can automatically maintain themselves throughout different scenarios. These systems have been described with many terms, such as self-healing, self-stabilizing, self-organizing, self-adaptive, self-optimizing, self-protecting, and self-managing. These attributes are collectively referred to as self-* properties. Even with the increased focus on self-* research, there exists much ambiguity in the perceptions of the different self-* properties. In this paper, we propose to resolve the ambiguity by introducing a template for defining self-* properties, and use it to offer formal definitions of existing self-* terms. We then present some observations about the relationships among the different self-* properties. Finally, we propose two new self-* properties that are meaningful in this space.

Super-fast distributed algorithms for metric facility location

This paper presents a distributed O(1)-approximation algorithm in the

Brief announcement: a framework for building self-stabilizing overlay networks

\mathcal{CONGEST}

Brief announcement: optimal self-stabilizing multi-token ring: a randomized solution

model for the metric facility location problem on a size-n clique network that has an expected running time of O(loglogn ·log*n) rounds. Though metric facility location has been considered by a number of researchers in low-diameter settings, this is the first sub-logarithmic-round algorithm for the problem that yields an O(1)-approximation in the setting of non-uniform facility opening costs. Since the facility location problem is specified by Ω(n2) bits of information, any fast solution in the

Avatar: A Time- and Space-Efficient Self-stabilizing Overlay Network

\mathcal{CONGEST}

Self-stabilizing Power-law Networks

model must be truly distributed. Our paper makes three main technical contributions. First, we show a new lower bound for metric facility location. Next, we demonstrate a reduction of the distributed metric facility location problem to the problem of computing an O(1)-ruling set of an appropriate spanning subgraph. Finally, we present a sub-logarithmic-round (in expectation) algorithm for computing a 2-ruling set in a spanning subgraph of a clique. Our algorithm accomplishes this by using a combination of randomized and deterministic sparsification.

Stabilizing pipelines for streaming applications

We describe a simple framework, called the transitive closure framework (TCF), for the self-stabilizing construction of any overlay network. The TCF is easy to reason about and algorithms derived from it stabilize within O(log n) more rounds than the optimal. As evidence of the power of this framework, we derive from the TCF a simple, self-stabilizing protocol for constructing Skip + graphs in O(log n) rounds.

A comparison of CORBA and ada's distributed systems annex

The token ring is a seminal topic in self-stabilization research. It has been expanded to include multiple tokens, and improved upon using randomization to lower the state space requirements. In this brief announcement, we discuss how the two ideas can be brought together for an optimal solution to the multi-token ring problem.

Building self-stabilizing overlay networks with the transitive closure framework

Overlay networks present an interesting challenge for fault-tolerant computing. Many overlay networks operate in dynamic environments e.g. the Internet, where faults are frequent and widespread, and the number of processes in a system may be quite large. Recently, self-stabilizing overlay networks have been presented as a method for managing this complexity. Self-stabilizing overlay networks promise that, starting from any weakly-connected configuration, a correct overlay network will eventually be built. To date, this guarantee has come at a cost: nodes may either have high degree during the algorithms execution, or the algorithm may take a long time to reach a legal configuration. In this paper, we present the first self-stabilizing overlay network algorithm that does not incur this penalty. Specifically, we i present a new locally-checkable overlay network based upon a binary search tree, and ii provide a randomized algorithm for self-stabilization that terminates in an expected polylogarithmic number of rounds and increases a nodes degree by only a polylogarithmic factor in expectation.

Searching for Malware in BitTorrent

Power-law graphs model the interconnections in various types of large-scale networks ranging from physical and biological systems to man-made social networks and web graphs. In these graphs, the degree distribution of the nodes obeys the power-law property: the fraction of nodes P(k) having a degree k closely follows the rule P(k) ∞ k−-γ. In the domain of man-made systems, if the topology of a power-law network gets altered due to failures or adversarial attacks, then remedial actions to restore the power-law property are very important. This paper presents self-stabilizing algorithms for maintaining the power-law property in a network of processes. These algorithms allow spontaneous restoration of the power-law property from any initial connected configuration. The algorithms consist of three modular components: a detection component to detect the violation of the power-law property, an interim topology creation component, and a repair component to build the final graph. We propose two different interim topologies, a clique and a linear graph. We then present two different techniques for rebuilding the power-law topology -- a probabilistic approach based on the preferential attachment model, which stabilizes in O(log n) communication rounds with a link complexity of O(n) per process, and a deterministic approach that introduces the novel data structure Bridge Tree and stabilizes in O(n) communication rounds with a much lower link complexity.