Exploiting Route Redundancy via Structured Peer to Peer Overlays
Exploiting Route Redundancy via Structured Peer to Peer Ben Y. Zhao, Ling Huang, Jeremy Stribling, Overlays Anthony D. Joseph, and John D. Kubiatowicz University of California, Berkeley ICNP 2003 Challenges Facing Network Applications Network connectivity is not reliable Disconnections frequent in the wide-area Internet IP-level repair is slow Wide-area: BGP 3 mins Local-area: IS-IS 5 seconds Next generation network applications Mostly wide-area Streaming media, VoIP, B2B transactions Low tolerance of delay, jitter and faults Our work: transparent resilient routing infrastructure that adapts to faults in not seconds, but milliseconds November 7, 2003
ICNP 2003 [email protected] Talk Overview Motivation Why structured routing Structured Peer to Peer overlays Mechanisms and policy Evaluation Summary November 7, 2003 ICNP 2003 [email protected] Routing in Mesh-like Networks Previous work has shown reasons for long convergence [Labovitz00, Labovitz01] MinRouteAdver timer Necessary to aggregate updates from all neighbors
Contributes to lower bound of BGP convergence time Internet becoming more mesh-like [Kaat99,labovitz99] Commonly set to 30 seconds Worsens BGP convergence behavior Question Can convergence be faster in context of structured routing? November 7, 2003 ICNP 2003 [email protected] Resilient Overlay Networks (MIT) Fully connected mesh Allows each node full knowledge of network
D Fast, independent calculation of routes Nodes can construct any path, maximum flexibility Cost of flexibility Protocol needs to choose the right route/nodes Per node O(n) state Monitors n - 1 paths O(n2) total path monitoring is expensive S November 7, 2003 ICNP 2003 [email protected] Leveraging Structured Peer-toPeer Overlays 0 source Key based routing (IPTPS 03)
Large sparse ID space N root(k) (160 bits: 0 2160) Nodes in overlay network have nodeIDs N Given some key k N, overlay deterministically maps k to its root node (live node in the network) route message to root (k) k Distributed Hashtables (DHT) is interface on KBR Key is leveraging underlying routing mesh November 7, 2003 ICNP 2003 [email protected] Proximity Neighbor Selection PNS = network aware overlay construction
Important for routing Within routing constraints, choose neighbors closest in network distance (latency) Generally reduces # of IP hops Reduce latency Reduce susceptibility to faults Less IP links = smaller chance of link/router failure Reduce overall network bandwidth utilization We use Tapestry to demonstrate our design P2P protocol with PNS overlay construction Topology-unaware P2P protocols will likely perform worse November 7, 2003 ICNP 2003 [email protected] System Architecture v v v v
v v OVERLAY v v v v v v v Internet Locate nearby overlay proxy Establish overlay path to destination host Overlay traffic routes traffic resiliently November 7, 2003 ICNP 2003 [email protected] Traffic Tunneling Legacy Node A A, B are IP addresses P(B)
B register register Proxy Legacy Node B Proxy P(A) = A get (hash(B)) P(B) put (hash(A), P(A)) P(B) = B put (hash(B), P(B)) Structured Peer to Peer Overlay Store mapping from end host IP to its proxys overlay ID Similar to approach in Internet Indirection Infrastructure (I3) November 7, 2003 ICNP 2003 [email protected] Tradeoffs of Tunneling via Less neighbor paths to monitor per node: O(log(n)) P2P
Actively maintain path redundancy Manageable for small # of paths Redirect traffic immediately when a failure is detected Eliminate on-the-fly calculation of new routes Restore redundancy when a path fails End result Large reduction in probing bandwidth: O(n) O(log(n)) Increase probing frequency Faster fault detection with low bandwidth consumption Fast fault detection + precomputed paths = increased responsiveness to faults Cons Overlay imposes routing stretch (more IP hops), generally < 2 November 7, 2003
ICNP 2003 [email protected] Some Details Efficient fault detection Use soft-state to periodically probe log(n) neighbor paths Small number of routes reduced bandwidth Exponentially weighted moving average in link quality estimation Avoid route flapping due to short term loss artifacts Loss rate Ln = (1 - ) Ln-1 + p p = instantaneous loss rate, = hysteresis factor Maintaining backup paths Each hop has flexible routing constraint Create and store backup routes at node insertion Restore redundancy via intelligent gossip after failures Simple policies to choose among redundant paths
November 7, 2003 ICNP 2003 [email protected] First Reachable Link Selection (FRLS) Use estimated loss results to choose shortest usable path Sort next hop paths by latency Use shortest path with 2299 minimal quality > T Correlated failures Reduce with intelligent topology construction Key is to leverage redundancy available November 7, 2003 ICNP 2003 2225 2274
2046 2281 2286 2530 1111 [email protected] Evaluation Metrics for evaluation How much routing resiliency can we exploit? How fast can we adapt to faults? What is the overhead of routing around a failure? Proportional increase in end to end latency Proportional increase in end to end bandwidth used Experimental platforms Event-based simulations on transit stub topologies
Data collected over different 5000-node topologies PlanetLab measurements Microbenchmarks on responsiveness Bandwidth measurements from 200+ node overlays Multiple virtual nodes run per physical machine November 7, 2003 ICNP 2003 [email protected] % of All Pairs Reachable Exploiting Route Redundancy 1 (Sim) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Instantaneous IP 0 0.05
Tapestry / FRLS 0.1 0.15 0.2 Proportion of IP Links Broken Simulation of Tapestry, 2 backup paths per routing entry Transit-stub topology shown, results from TIER and AS graphs similar November 7, 2003 ICNP 2003 [email protected] Responsiveness to Faults (PlanetLab) Time to Switch Routes (ms) 2500 2000 1500 1000 660 alpha=0.2 alpha=0.4
500 0 0 200 300 400 600 800 1000 1200 Link Probe Period (ms) Response time increases linearly with probe period Minimum link quality threshold T = 70%, 20 runs per data point November 7, 2003 ICNP 2003 [email protected] Link Probing Bandwidth (Planetlab) Bandwidth Per Node (KB/s) 7 PR=300ms
PR=600ms 6 5 4 3 2 1 0 1 10 100 1000 Size of Overlay Medium sized routing overlays incur low probing bandwidth Bandwidth increases logarithmically with overlay size November 7, 2003 ICNP 2003 [email protected] Related Work Redirection overlays
Can no longer construct arbitrary paths Structured routing with low redundancy gets very close to ideal in connectivity Incur low routing stretch Fast enough for highly interactive applications Allows caching of backup paths for quick failover 300ms beacon period response time < 700ms On overlay networks of 300 nodes, b/w cost is 7KB/s Future work Deploying a public routing and proxy service on PlanetLab Examine impact of Network aware topology construction Loss sensitive probing techniques November 7, 2003
ICNP 2003 [email protected] Questions Related websites: Tapestry http://www.cs.berkeley.edu/~ravenben/tapestry Pastry http://research.microsoft.com/~antr/pastry Chord http://lcs.mit.edu/chord Acknowledgements Thanks to Dennis Geels and Sean Rhea for their work on the BMark benchmark suite November 7, 2003 ICNP 2003 [email protected]
Machu Picchu, Peru The Andes Mountains run through entire west coast of South America, from Colombia to Chile. Length: 4,350 miles The Amazon River runs into the north of South America, through Brazil, Colombia, Ecuador, and Peru. Length: 4,200 miles...
that evoke sympathy or distaste. Among the . strongest emotional appeals are anecdotes or examples that dramatize a situation. For instance, Paine's story of the Tory tavern keeper and his nine-year-old child makes a strong appeal to the human desire...
Cost effective - Can be distributed to lots of people (= large amounts of data) and the researcher need not be present.. Easy to analyse . if fixed choice, closed questions are used (statistical analysis and comparisons) X Social desirability...
9-Line Evacuation Request. Line 5: Number of litter or ambulatory. Said as L-#, A-#. Line 6 tells evacuation control about the enemy situation near the evacuation point, and whether escort is needed. Often, lines 1-5 and/or 6 are enough information...
Weather Maps Weather Maps Temp dewpt pressure trend change shown low clouds middle clouds high clouds wxnow visibility low or middle cloud cover PAST when base amount The amount of precipitation Amount of precipitation is plotted in inches to the...
Relief Sculpture Ancient Egyptian Pottery Pottery was used by the ancient Egyptians in much the same way we use modern kitchen containers or plastic, Two distinct Types Nile silt ware - Nile clay. After being fired, it has a red-brown...
Film as Ethnography; or, Translation between Cultures in the Postcolonial World Group One Chow discusses the issue of translation by using Benjamin's theory: "The Task of the Translator" Translation and the problem of Origins Our discussion here can be facilitated...
Mining the data warehouse. ... A leader in premium pizza market. 130 full-service restaurants. 26 states, D.C., and 5 foreign countries. ... Illustrate why 100 percent accurate and complete information is impossible for Noodles & Company to obtain.
Ready to download the document? Go ahead and hit continue!