Citation Context

...ired by the De-Bruijn graph. We are not the only ones to use the DeBruijn graph in this context. Constructions using it were suggested independently by Fraigniaud and Gauron [12], Kaashoek and Karger =-=[18]-=- and Abraham et al.[1]. The parameters they achieve are similar to those proved in Section 2, yet their approach in emulating the De-Bruijn graph is different. They try to do it directly whereas we us...

Citation Context

...] uses skip-lists to provide O(log N) probabilistic lookup guarantee. Viceroy [23] provides O(log N) lookup 341 performance with a constant degree routing graph. A few log N DHTs use De Bruijn graphs =-=[21, 40]-=- to achieve O( log log N ) lookup performance. The multi-hop lookup performance provided by these DHTs is inadequate to support performance sensitive application like DNS. A few recent DHTs provide O(...

Citation Context

...r of nodes. We will compare two different ways of selecting long-distance neighbors. Deterministic and randomized topologies In the first class of topologies, used in Chord [34], CAN [26], and others =-=[17, 25]-=-, each node v maintains links to the owners of certain other IDs which are a deterministic function of v’s ID. For example, Chord’s keyspace is {0,...,N − 1}, whereN = 2 160 , and node v maintains lin...

Citation Context

...gement. Recent years have seen a large number of structured overlays that organize the network based on rings [33, 29, 39, 23], hyperdimensional cubes [28], butterfly structures [22], deBruijn graphs =-=[17, 38]-=-, or skip-lists [14]. Corona is agnostic about the choice of the overlay and can be easily layered on any overlay with uniform node degree, including the ones listed here. Corona’s approach to a peer-...

Citation Context

...Gnutella, Napster, Kazaa, Morpheus, BitTorrent, and many others. Several distributed hash tables (DHTs) have been introduced which are provably robust to random peer deletions (i.e. fail-stop faults) =-=[1, 13, 16, 21, 22, 26, 28]-=-. We are aware of several results that deal with the more challenging problem of designing DHTs which are robust to Byzantine faults. All of these results are robust only to the random Byzantine attac...

Citation Context

...Figure 5. IV. ANALYSIS In this section we analyze the performance of Y0 with respect to the metrics described in Section II-B. The results apply Y0’s techniques to any ring-based DHT, e.g [33], [29], =-=[17]-=-, [26], [23]. We prove the following: • Y0 can achieve a near-optimal maximum share of 1 + ε for any ε > 0 (Section IV-A). • As long as the number of nodes n and the average capacity remain relatively...

Citation Context

...is primarily designed to address the first two: query latency and load balance. A query is merely an instance of data location (a lookup) with a fully qualified name. A number of efficient algorithms =-=[26, 18, 14, 21, 20, 4, 15, 13]-=- provide low average query latency, and we do not consider it here further. Distributing load equitably, however, is more difficult. For example, many recent systems [26, 28, 20, 21] attempt to balanc...

Citation Context

...th an example run. Figure 2: The filtered limit distribution (si) after clustering 5.1 Algorithm Example Let us consider a specific application of our algorithm on a synthetically-generated de Bruijn =-=[41]-=- peer-to-peer graph embedded within a communication graph sampled from the Internet (using NetFlow traces from the Abilene Internet2 ISP). The Abilene communication graph GD contains |VD| = 104426 nod...

Citation Context

...of the peers in the system at a certain time P (P ⊆ I). To make the rest of the paper concrete, we will define a distance function, as well as a mapping from identifiers to peers, as commonly used in =-=[2, 1, 3, 4]-=-. Our replication scheme, however, does not assume these definitions and can thus be applied to a variety of structured P2P systems. We will assume the distance function is defined as: d(x, y) = y ⊖ x...