Results 1  10
of
22
Know thy Neighbor's Neighbor: the Power of Lookahead in Randomized P2P Networks
 In Proceedings of the 36th ACM Symposium on Theory of Computing (STOC
, 2004
"... Several peertopeer networks are based upon randomized graph topologies that permit e#cient greedy routing, e.g., randomized hypercubes, randomized Chord, skipgraphs and constructions based upon smallworld percolation networks. In each of these networks, a node has outdegree #(log n), where n de ..."
Abstract

Cited by 102 (5 self)
 Add to MetaCart
Several peertopeer networks are based upon randomized graph topologies that permit e#cient greedy routing, e.g., randomized hypercubes, randomized Chord, skipgraphs and constructions based upon smallworld percolation networks. In each of these networks, a node has outdegree #(log n), where n denotes the total number of nodes, and greedy routing is known to take O(log n) hops on average. We establish lowerbounds for greedy routing for these networks, and analyze NeighborofNeighbor (NoN)greedy routing. The idea behind NoN, as the name suggests, is to take a neighbor's neighbors into account for making better routing decisions.
Distributed placement of service facilities in largescale networks
, 2006
"... Abstract — The effectiveness of service provisioning in largescale networks is highly dependent on the number and location of service facilities deployed at various hosts. The classical, centralized approach to determining the latter would amount to formulating and solving the uncapacitated kmedian ..."
Abstract

Cited by 23 (3 self)
 Add to MetaCart
(Show Context)
Abstract — The effectiveness of service provisioning in largescale networks is highly dependent on the number and location of service facilities deployed at various hosts. The classical, centralized approach to determining the latter would amount to formulating and solving the uncapacitated kmedian (UKM) problem (if the requested number of facilities is fixed), or the uncapacitated facility location (UFL) problem (if the number of facilities is also to be optimized). Clearly, such centralized approaches require knowledge of global topological and demand information, and thus do not scale and are not practical for large networks. The key question posed and answered in this paper is the following: “How can we determine in a distributed and scalable manner the number and location of service facilities?” We propose an innovative approach in which topology and demand information is limited to neighborhoods, or balls of small radius around selected facilities, whereas demand information is captured implicitly for the remaining (remote) clients outside these neighborhoods, by mapping them to clients on the edge of the neighborhood; the ball radius regulates the tradeoff between scalability and performance. We develop a scalable, distributed approach that answers our key question through an iterative reoptimization of the location and the number of facilities within such balls. We show that even for small values of the radius (1 or 2), our distributed approach achieves performance under various synthetic and real Internet topologies that is comparable to that of optimal, centralized approaches requiring full topology and demand information.
Skipwebs: Efficient distributed data structures for multidimensional data sets
 In 24th ACM Symp. on Principles of Distributed Computing (PODC
, 2005
"... large(at)daimi.au.dk eppstein(at)ics.uci.edu goodrich(at)acm.org We present a framework for designing efficient distributed data structures for multidimensional data. Our structures, which we call skipwebs, extend and improve previous randomized distributed data structures, including skipnets and ..."
Abstract

Cited by 19 (0 self)
 Add to MetaCart
(Show Context)
large(at)daimi.au.dk eppstein(at)ics.uci.edu goodrich(at)acm.org We present a framework for designing efficient distributed data structures for multidimensional data. Our structures, which we call skipwebs, extend and improve previous randomized distributed data structures, including skipnets and skip graphs. Our framework applies to a general class of data querying scenarios, which include linear (onedimensional) data, such as sorted sets, as well as multidimensional data, such as ddimensional octrees and digital tries of character strings defined over a fixed alphabet. We show how to perform a query over such a set of n items spread among n hosts using O(log n/log log n) messages for onedimensional data, or O(log n) messages for fixeddimensional data, while using only O(log n) space per host. We also show how to make such structures dynamic so as to allow for insertions and deletions in O(log n) messages for quadtrees, octrees, and digital tries, and O(log n/log log n) messages for onedimensional data. Finally, we show how to apply a blocking strategy to skipwebs to further improve message complexity for onedimensional data when hosts can store more data.
Wired geometric routing
 In Proceedings of IPTPS 2007
, 2007
"... Routing substrates for overlay networks are an important building block for large distributed applications. Many existing substrates are based on a random identifier space and therefore do not respect node locality when routing data. This can lead to lower performance for localitysensitive applicat ..."
Abstract

Cited by 19 (2 self)
 Add to MetaCart
(Show Context)
Routing substrates for overlay networks are an important building block for large distributed applications. Many existing substrates are based on a random identifier space and therefore do not respect node locality when routing data. This can lead to lower performance for localitysensitive applications, such as web caching, distributed gaming, and resource discovery. This paper examines the problem of building a localityaware routing substrate on top of a localitybased coordinate system, where the distance between coordinates approximates network latencies. As a starting point we take the scaled θrouting proposal for geometric routing in a Euclidean space. We address the practical problems of forming routing tables with imperfect node knowledge and churn and examine query performance on nonEuclidean data sets. 1
The Rainbow Skip Graph: A FaultTolerant ConstantDegree Distributed Data Structure
"... We present a distributed data structure, which we call the rainbow skip graph. To our knowledge, this is the first peertopeer data structure that simultaneously achieves high faulttolerance, constantsized nodes, and fast update and query times for ordered data. It is a nontrivial adaptation of ..."
Abstract

Cited by 17 (0 self)
 Add to MetaCart
We present a distributed data structure, which we call the rainbow skip graph. To our knowledge, this is the first peertopeer data structure that simultaneously achieves high faulttolerance, constantsized nodes, and fast update and query times for ordered data. It is a nontrivial adaptation of the SkipNet/skipgraph structures of Harvey et al. and Aspnes and Shah, so as to provide faulttolerance as these structures do, but to do so using constantsized nodes, as in the family tree structure of Zatloukal and Harvey. It supports successor queries on a set of n items using O(log n) messages with high probability, an improvement over the expected O(log n) messages of the family tree. Our structure achieves these results by using the following new constructs: • Rainbow connections: parallel sets of pointers between related components of nodes, so as to achieve good connectivity between “adjacent ” components, using constantsized nodes. • Hydra components: highlyconnected, highly faulttolerant components of constantsized nodes, which will contain relatively large connected subcomponents even under the failure of a constant fraction of the nodes in the component. We further augment the hydra components in the rainbow skip graph by using erasureresilient codes to ensure that any large subcomponent of nodes in a hydra component is sufficient to reconstruct all the data stored in that component. By carefully maintaining the size of related components and hydra components to be O(log n), we are able to achieve fast times for updates and queries in the rainbow skip graph. In addition, we show how to make the communication complexity for updates and queries be worst case, at the expense of more conceptual complexity and a slight degradation in the node congestion of the data structure.
Efficient Content Authentication over Distributed Hash Tables
, 2005
"... We study a new model for data authentication over peertopeer storage networks, where data is stored, queried and authenticated in a totally distributed fashion. The model captures the security requirements of emerging distributed computing applications. We present an efficient implementation of a ..."
Abstract

Cited by 8 (0 self)
 Add to MetaCart
We study a new model for data authentication over peertopeer storage networks, where data is stored, queried and authenticated in a totally distributed fashion. The model captures the security requirements of emerging distributed computing applications. We present an efficient implementation of a distributed Merkle tree, which realizes a Merkle tree over a peertopeer network, thus extending a fundamental cryptographic authentication technique to a peertopeer distributed environment. We show how our distributed Merkle tree can be used to design an efficient authenticated distributed hash table. Our scheme is built on top of a broad class of existing distributed hash table implementations, is efficient, and achieves generality by only using the basic functionality of object location. We use this scheme to implement the first distributed authenticated dictionary.
Picking up the Pieces: SelfHealing in Reconfigurable Networks
"... We consider the problem of selfhealing in networks that are reconfigurable in the sense that they can change their topology during an attack. Our goal is to maintain connectivity in these networks, even in the presence of repeated adversarial node deletion, by carefully adding edges after each atta ..."
Abstract

Cited by 8 (3 self)
 Add to MetaCart
(Show Context)
We consider the problem of selfhealing in networks that are reconfigurable in the sense that they can change their topology during an attack. Our goal is to maintain connectivity in these networks, even in the presence of repeated adversarial node deletion, by carefully adding edges after each attack. We present a new algorithm, DASH, that provably ensures that: 1) the network stays connected even if an adversary deletes up to all nodes in the network; and 2) no node ever increases its degree by more than 2 log n, where n is the number of nodes initially in the network. DASH is fully distributed; adds new edges only among neighbors of deleted nodes; and has average latency and bandwidth costs that are at most logarithmic in n. DASH has these properties irrespective of the topology of the initial network, and is thus orthogonal and complementary to traditional topologybased approaches to defending against attack. We also prove lowerbounds showing that DASH is asymptotically optimal in terms of minimizing maximum degree increase over multiple attacks. Finally, we present empirical results on powerlaw graphs that show that DASH performs well in practice, and that it significantly outperforms naive algorithms in reducing maximum degree increase.
Distributed Server Migration for Scalable Internet Service Deployment
 SUBMITTED TO IEEE/ACM TRANSACTIONS ON NETWORKING MAY/XX/2008
, 2008
"... The effectiveness of service provisioning in largescale networks is highly dependent on the number and location of service facilities deployed at various hosts. The classical, centralized approach to determining the latter would amount to formulating and solving the uncapacitated kmedian (UKM) pro ..."
Abstract

Cited by 6 (4 self)
 Add to MetaCart
The effectiveness of service provisioning in largescale networks is highly dependent on the number and location of service facilities deployed at various hosts. The classical, centralized approach to determining the latter would amount to formulating and solving the uncapacitated kmedian (UKM) problem (if the requested number of facilities is fixed), or the uncapacitated facility location (UFL) problem (if the number of facilities is also to be optimized). Clearly, such centralized approaches require knowledge of global topological and demand information, and thus do not scale and are not practical for large networks. The key question posed and answered in this paper is the following: “How can we determine in a distributed and scalable manner the number and location of service facilities?” We propose an innovative approach in which topology and demand information is limited to neighborhoods, or balls of small radius around selected facilities, whereas demand information is captured implicitly for the remaining (remote) clients outside these neighborhoods, by mapping them to clients on the edge of the neighborhood; the ball radius regulates the tradeoff between scalability and performance. We develop a scalable, distributed approach that answers our key question through an iterative reoptimization of the location and the number of facilities within such balls. We show that even for small values of the radius (1 or 2), our distributed approach achieves performance under various synthetic and real Internet topologies and workloads that is comparable to that of optimal, centralized approaches requiring full topology and demand information.
DegreeOptimal Deterministic Routing for P2P Systems
 In Proceedings of 10th IEEE Symposium on computers and communications (ISCC ’05) La Manga del Mar Menor
, 2005
"... We propose routing schemes that optimize the average number of hops for lookup requests in Peer–to–Peer (P2P) systems without adding any overhead to the system. Our work is inspired by the recently introduced variation of greedy routing, called neighbor–of–neighbor (NoN), which allows to get optimal ..."
Abstract

Cited by 6 (4 self)
 Add to MetaCart
(Show Context)
We propose routing schemes that optimize the average number of hops for lookup requests in Peer–to–Peer (P2P) systems without adding any overhead to the system. Our work is inspired by the recently introduced variation of greedy routing, called neighbor–of–neighbor (NoN), which allows to get optimal average path length with respect to the degree. Our proposal has the advantage of first “limiting” and then “eliminating ” the use of randomization. As a consequence, the NoN technique can be implemented with our schemes without adding any overhead. Analyzed networks include several popular topologies: Chord, Hypercube based networks, Symphony, SkipGraphs. Theoretical results and extensive simulations show that the proposed simplifications (while maintaining the original node degree) do not increase the average path length of the networks, which is often improved in practice. The improvement is obtained with no harm to the operational efficiency (e.g. stability, ease of programming, scalability, fault–tolerance) of the considered systems. 1
FChord: Improved uniform routing on Chord
 Proc. 11th Colloquium on Structural Information and Communication Complexity
, 2004
"... We propose a family of novel Chordbased P2P schemes retaining all positive aspects that made Chord a popular topology for routing in P2P networks. The schemes, based on the Fibonacci number system, allow to simultaneously improve on the maximum/average number of hops for lookups and the routing tab ..."
Abstract

Cited by 5 (3 self)
 Add to MetaCart
(Show Context)
We propose a family of novel Chordbased P2P schemes retaining all positive aspects that made Chord a popular topology for routing in P2P networks. The schemes, based on the Fibonacci number system, allow to simultaneously improve on the maximum/average number of hops for lookups and the routing table size per node.