Browsing by Author "Haeberlen, Andreas"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Accountability for distributed systems
    (2009) Haeberlen, Andreas; Druschel, Peter
    Nodes in a distributed system can fail for many reasons, such as bugs, misconfigurations, hardware failures, intrusions, or insider attacks. Once a node has become faulty, its behavior can change arbitrarily. In benign cases, the node might simply stop; in less benign cases, it might actively try to subvert the rest of the system. A reliable distributed system must have a way to handle such faults. In this thesis, we explore a novel approach to this problem, which is based on accountability. In an accountable system, each node records its past actions in a tamper-evident log, and nodes inspect each other's log for signs of misbehavior. When nodes become faulty, the other nodes can eventually detect this, and they can obtain evidence that irrefutably links the fault to a faulty node. At the same time, correct nodes can always defend themselves against any false accusations. We characterize the class of faults that can be detected with our approach, and we show that it includes any fault that causally affects at least one correct node. We also present a set of techniques for enforcing accountability, including an algorithm for tamper-evident logs, and two techniques for detecting faults in the log: One relies on state machine replay to check a node's behavior against a reference implementation, while the other checks the logs against a declarative specification of the expected behavior. Each of these techniques can be applied to a wide range of distributed systems. To demonstrate that accountability is widely applicable, we have added it to several different types of systems, including a decentralized email system, a server-based file system, a peer-to-peer content distribution system, the Internet's interdomain routing system, and two multi-player games. In each case, accountability was able to detect a variety of problems that were previously reported in the literature. This shows that accountability is very general and can supersede a number of existing defenses. Our evaluation shows that accountability is practical, that its overhead is reasonable, and that it can scale to large numbers of nodes.
  • Thumbnail Image
    Item
    Consistent Key Mapping in Structured Overlays
    (2005-08-16) Druschel, Peter; Haeberlen, Andreas; Hoye, Jeff; Mislove, Alan
    Most structured peer-to-peer overlays rely on consistent hashing to determine the node that is responsible for a given key. For consistent hashing to work properly, it is necessary that the nodes have a consistent view of their neighborhood in the identifier space. However, if routing anomalies occur in the underlying network, this view can become inconsistent, causing unstable overlay behavior and, worse, allowing more than one node to assume responsibility for ranges of keys. We present a set of techniques for preventing inconsistencies under routing anomalies, and we propose to adopt strategies from mobile ad-hoc networking for maintaining connectivity in the presence of path failures. We evaluate our design in the context of Pastry and present results from a deployment in the PlanetLab testbed.