The Impact of DHT Routing Geometry on Resilience and Proximity K. Gummadi, R. Gummadi..,S.Gribble, S. Ratnasamy, S. Shenker, I. Stoica.

Slides:



Advertisements
Similar presentations
SkipNet: A Scalable Overlay Network with Practical Locality Properties Nick Harvey, Mike Jones, Stefan Saroiu, Marvin Theimer, Alec Wolman Microsoft Research.
Advertisements

CAN 1.Distributed Hash Tables a)DHT recap b)Uses c)Example – CAN.
1 S4: Small State and Small Stretch Routing for Large Wireless Sensor Networks Yun Mao 2, Feng Wang 1, Lili Qiu 1, Simon S. Lam 1, Jonathan M. Smith 2.
Ion Stoica, Robert Morris, David Karger, M. Frans Kaashoek, Hari Balakrishnan MIT and Berkeley presented by Daniel Figueiredo Chord: A Scalable Peer-to-peer.
Scalable Content-Addressable Network Lintao Liu
1 Greedy Forwarding in Dynamic Scale-Free Networks Embedded in Hyperbolic Metric Spaces Dmitri Krioukov CAIDA/UCSD Joint work with F. Papadopoulos, M.
Kademlia: A Peer-to-peer Information System Based on the XOR Metric Petar Mayamounkov David Mazières A few slides are taken from the authors’ original.
1 PASTRY Partially borrowed from Gabi Kliot ’ s presentation.
Common approach 1. Define space: assign random ID (160-bit) to each node and key 2. Define a metric topology in this space,  that is, the space of keys.
Presented by Elisavet Kozyri. A distributed application architecture that partitions tasks or work loads between peers Main actions: Find the owner of.
Scribe: A Large-Scale and Decentralized Application-Level Multicast Infrastructure Miguel Castro, Peter Druschel, Anne-Marie Kermarrec, and Antony L. T.
1 Canon in G Major: Designing DHTs with Hierarchical Structure Prasanna Ganesan (Stanford University) Krishna Gummadi (U. of Washington) Hector Garcia-Molina.
The Impact of DHT Routing Geometry on Resilience and Proximity New DHTs constantly proposed –CAN, Chord, Pastry, Tapestry, Plaxton, Viceroy, Kademlia,
Topics in Reliable Distributed Systems Lecture 2, Fall Dr. Idit Keidar.
SkipNet: A Scalable Overlay Network with Practical Locality Properties Nick Harvey, Mike Jones, Stefan Saroiu, Marvin Theimer, Alec Wolman Presented by.
Looking Up Data in P2P Systems Hari Balakrishnan M.Frans Kaashoek David Karger Robert Morris Ion Stoica.
SkipNet Christian Schmidt-Madsen, Peter Tiedemann,
P2P: Advanced Topics Filesystems over DHTs and P2P research Vyas Sekar.
Before start… Earlier work single-path routing in sensor networks
The Impact of DHT Routing Geometry on Resilience and Proximity Krishna Gummadi, Ramakrishna Gummadi, Sylvia Ratnasamy, Steve Gribble, Scott Shenker, Ion.
SkipNet: A Scalable Overlay Network with Practical Locality Properties Nick Harvey, Mike Jones, Stefan Saroiu, Marvin Theimer, Alec Wolman Microsoft Research.
Distributed Lookup Systems
University of Oregon Slides from Gotz and Wehrle + Chord paper
The Impact of DHT Routing Geometry on Resilience and Proximity Krishna Gummadi, Ramakrishna Gummadi, Sylvia Ratnasamy, Steve Gribble, Scott Shenker, Ion.
A Note on Finding the Nearest Neighbor in Growth-Restricted Metrics Kirsten Hildrum John Kubiatowicz Sean Ma Satish Rao.
1 Canon in G Major: Designing DHTs with Hierarchical Structure Prasanna Ganesan (Stanford University) Krishna Gummadi (U. of Washington) Hector Garcia-Molina.
1 Lecture 24: Interconnection Networks Topics: communication latency, centralized and decentralized switches (Sections 8.1 – 8.5)
SkipNet: A Scaleable Overlay Network With Practical Locality Properties Presented by Rachel Rubin CS294-4: Peer-to-Peer Systems By Nicholas Harvey, Michael.
Topics in Reliable Distributed Systems Fall Dr. Idit Keidar.
Decentralized Location Services CS273 Guest Lecture April 24, 2001 Ben Y. Zhao.
Or, Providing Scalable, Decentralized Location and Routing Network Services Tapestry: Fault-tolerant Wide-area Application Infrastructure Motivation and.
File Sharing : Hash/Lookup Yossi Shasho (HW in last slide) Based on Chord: A Scalable Peer-to-peer Lookup Service for Internet ApplicationsChord: A Scalable.
Internet Indirection Infrastructure (i3) Ion Stoica, Daniel Adkins, Shelley Zhuang, Scott Shenker, Sonesh Surana UC Berkeley SIGCOMM 2002.
 Structured peer to peer overlay networks are resilient – but not secure.  Even a small fraction of malicious nodes may result in failure of correct.
Mobile Ad-hoc Pastry (MADPastry) Niloy Ganguly. Problem of normal DHT in MANET No co-relation between overlay logical hop and physical hop – Low bandwidth,
Distributed Hash Table Systems Hui Zhang University of Southern California.
1 PASTRY. 2 Pastry paper “ Pastry: Scalable, decentralized object location and routing for large- scale peer-to-peer systems ” by Antony Rowstron (Microsoft.
PIC: Practical Internet Coordinates for Distance Estimation Manuel Costa joint work with Miguel Castro, Ant Rowstron, Peter Key Microsoft Research Cambridge.
DHTs and Peer-to-Peer Systems Supplemental Slides Aditya Akella 03/21/2007.
Impact of Neighbor Selection on Performance and Resilience of Structured P2P Networks IPTPS Feb. 25, 2005 Byung-Gon Chun, Ben Y. Zhao, and John Kubiatowicz.
Information-Centric Networks07a-1 Week 7 / Paper 1 Internet Indirection Infrastructure –Ion Stoica, Daniel Adkins, Shelley Zhuang, Scott Shenker, Sonesh.
Canon in G Major: Designing DHTs with Hierarchical Structure Prasanna Ganesan(Stanford University) Krishna Gummadi(University of Washington) Hector Garcia-Molina(Stanford.
Canon in G Major: Designing DHTs with Hierarchical Structure Prasanna Ganesan, Krishna Gummadi, Hector Garcia-Molina ICDCS2004 報告者 : 陳奕鈞.
Ion Stoica, Robert Morris, David Karger, M. Frans Kaashoek, Hari Balakrishnan MIT and Berkeley presented by Daniel Figueiredo Chord: A Scalable Peer-to-peer.
An IP Address Based Caching Scheme for Peer-to-Peer Networks Ronaldo Alves Ferreira Joint work with Ananth Grama and Suresh Jagannathan Department of Computer.
Chord: A Scalable Peer-to-peer Lookup Service for Internet Applications Ion Stoica, Robert Morris, David Karger, M. Frans Kaashoek, Hari Balakrishnan Presented.
Paper Survey of DHT Distributed Hash Table. Usages Directory service  Very little amount of information, such as URI, metadata, … Storage  Data, such.
Peer to Peer A Survey and comparison of peer-to-peer overlay network schemes And so on… Chulhyun Park
P2P Group Meeting (ICS/FORTH) Monday, 28 March, 2005 A Scalable Content-Addressable Network Sylvia Ratnasamy, Paul Francis, Mark Handley, Richard Karp,
Lecture 12 Distributed Hash Tables CPE 401/601 Computer Network Systems slides are modified from Jennifer Rexford.
Plethora: Infrastructure and System Design. Introduction Peer-to-Peer (P2P) networks: –Self-organizing distributed systems –Nodes receive and provide.
A Comparative Study of the DNS Design with DHT-Based Alternatives 95/08/31 Chen Chih-Ming.
CS 268: Computer Networking
Topologically-Aware Overlay Construction and Sever Selection Sylvia Ratnasamy, Mark Handley, Richard Karp, Scott Shenker.
Peter R Pietzuch and Jean Bacon Peer-to-Peer Overlay Networks in an Event-Based Middleware DEBS’03, San Diego, CA, USA,
LOOKING UP DATA IN P2P SYSTEMS Hari Balakrishnan M. Frans Kaashoek David Karger Robert Morris Ion Stoica MIT LCS.
Spring 2000CS 4611 Routing Outline Algorithms Scalability.
Data Structure of Chord  For each peer -successor link on the ring -predecessor link on the ring -for all i ∈ {0,..,m-1} Finger[i] := the peer following.
CS694 - DHT1 Distributed Hash Table Systems Hui Zhang University of Southern California.
Peer-to-Peer Protocol (P2PP) Salman Baset, Henning Schulzrinne Columbia University.
Incrementally Improving Lookup Latency in Distributed Hash Table Systems Hui Zhang 1, Ashish Goel 2, Ramesh Govindan 1 1 University of Southern California.
Distributed Hash Tables
Impact of Neighbor Selection on Performance and Resilience of Structured P2P Networks Sushma Maramreddy.
Plethora: Infrastructure and System Design
PASTRY.
DHT Routing Geometries and Chord
Interconnection Network Design Lecture 14
COS 461: Computer Networks
Exploiting Routing Redundancy via Structured Peer-to-Peer Overlays
P2P: Distributed Hash Tables
Presentation transcript:

The Impact of DHT Routing Geometry on Resilience and Proximity K. Gummadi, R. Gummadi..,S.Gribble, S. Ratnasamy, S. Shenker, I. Stoica

Introduction how routing geometries affect the resilience and proximity properties of DHTs –coping with node failures: static resilience –adapting to Internet topology, proximity issues: path latency and local convergence –flexibility in selection of neighbors and routes an essential factor evaluating various geometries for gaining insight for better designs admitted flaws –not evaluating all routing algorithms –not considering management overhead –focusing on only two performance issues

Terminology Algorithm: the exact details of selecting neighbors and next-hops Geometry: a geometric interpretation of how the selections are made –geometry constrains the choices, but small changes in algorithm do not change the geometry Flexibility: the algorithmic freedom left after the geometry is chosen –neighbor selection does the geometry allow selecting neighbors based on proximity does the geometry support sequential neighbors –route selection how many options for next-hop in case of failure does the geometry allow selecting next-hops based on proximity Sequential neighbors –can route and progress towards all destinations –global ordering on distances required (naturally in Ring, add-on for others)

Basic Routing Geometries (1/2) Tree -e.g. PRR -node ids are leaf nodes -distance = smallest common subtree -routing MSB first Hypercube -e.g. CAN -node id represents position -neighbors differ by 1 bit only -distance is # of differing bits -routing in any order, previous corrections preserved Butterfly -e.g. Viceroy -nodes in log n stages -nodes at stage i can correct i th bit -imposed global ordering: global and stage successors/predecessors required to be held as neighbors -routing in log n hops with constant state at each node: first through stages, then through successors/predecessors

Basic Routing Geometries (2/2) Ring -e.g. Chord -nodes in one-dimensional cyclic id space -distance is the clockwise numeric distance between nodes -routing in log n hops, because each hop cuts the distance in half XOR -e.g. Kademlia -distance is XOR of ids -routing MSB first, but in case of failures can correct the next bit -these corrections are not necessarily preserved => multiple non-optimal paths Hybrid -e.g. Pastry -dual mode, e.g. tree + ring -nodes both leafs and on circle -distance both tree and ring distance -fallback to ring when tree fails -can progress on the tree while not on progressing on the ring

Static Resilience DHTs resilient to node failures Three aspects of resilience –data replication –routing recovery –static resilience routing before recovery algorithms kick in the only aspect considered in this paper

Static Resilience Performance of DHTs with routing tables of equal sizes What happens when failures are present? –failed paths? –increase in path length?

Static Resilience of Various Geometries consistent with flexibility in route selection Ring and Hypercube have twice the flexibility of Hybrid and XOR Tree and Butterfly have no flexibility Hypercube has equal length alternative paths Hybrid, Tree, and XOR have only longer alternative paths Ring has some longer paths Butterfly not applicable

Static Resilience: Sequential Neighbors 16 sequential neighbors added XOR not applicable, Tree included in Hybrid Ring has natural support for sequential neighbors greatly increased resilience increase in resilience comes at the cost of path stretch note the scale! path increase in Butterfly is way higher

Static Resilience: Sequential vs. Regular Neighbors at high failure rates, sequentials are better sequentials can lead to longer paths a large number of regulars is a good compromise

Proximity DHTs designed to route effectively in terms of hopcounts end-to-end latency issues approached through proximity methods PNS - neighbor selection - identifying closest nodes is hard => heuristics needed - PNS(K), K=# of samples - a node’s latency distribution helps in choosing K PRS - next-hop selection -complicated tradeoff between # of hops and latency - different heuristics for different geometries PIS - ids based on location - load balancing hard - not discussed here

Proximity: Performance Results Basic assumptions in this evaluation –evaluation based on recursive routing –geometries have little effect, support for PNS/PRS essential –performance of proximity methods depends on topology and its latency characteristics good approximation by looking at the latency distribution for a typical node –Hybrid left out, Butterfly not applicable –PNS/PRS not expensive in terms of hopcounts

Proximity: PNS vs. PRS XOR and Ring support both PRS better than plain PNS far better (PNS 2 i neighbor options, PRS i options) PRS+PNS only a small improvement over PNS PNS(16) gives similar results 16 sequential neighbors improve PRS and Plain a little, but not significantly => PNS support most important

Proximity: More Results Does geometry matter? –Tree: PRS n/a –Hypercube: PNS n/a –performance of pairs very close => PNS/PRS support matters, not geometry Absolute performance –depends on latency distribution => overlay latencies can be reduced to small multiples of underlying Internet path latencies

Local Convergence (1/2) messages sent from two nearby nodes converge at a node near the sources => low latencies or bandwidth savings in e.g. - overlay multicast - caching - server selection measured by number of exit points PNS vs. PRS: PNS nearly optimal, PRS not effective

Local Convergence (2/2) PNS(16) and PRS equally ineffective PNS(16)+PRS better => PRS more relevant with limited samples and domain sizes => geometries don’t matter, PNS/PRS support is the key

Summary geometry constrains other design choices flexibility important (neighbor & next-hop selection) Ring & XOR are flexible: support for both PNS and PRS Why not Ring? –flexibility –natural support for sequential neighbors –tested well