Topology-Aware Overlay Construction and Server Selection

1
Topology-Aware Overlay Construction and Server
Selection

• Sylvia Ratnasamy
• Mark Handley
• Richard Karp
• Scott Shenker

Infocom 2002
2
Connections of a node
3
Introduction

• Problem Inefficient routing in large-scale
  networks
• In large-scale overlay networks, each node is
  logically connected to a small subset of other
  participants.
• Due to the lack of effort to ensure that
  application-level connectivity is congruent with
  underlying IP-level network topology
• Basic Idea Optimize routing paths in network
• Define a binning scheme whereby nodes partition
  themselves into bins
• Nodes that fall within a given bin are relatively
  close to one another in terms of network latency

4
Outline

• Introduction
• Distributed Binning
• Topologically-aware construction of overlay
  networks
• Topologically-aware server selection
• Conclusion

5
Extracting proximity information

• Measuments that can be used to derive topological
  information
• traceroute ?
• intended for network diagnostic purposes,
• too heavy-weight,
• excessive load on the network,
• disabled ICMP at some sites for security
• BGP routing table ?
• not directly available for end users,
• requires privilege or third party service
• Network latency ?
• often a direct indicator of network performance,
• light-weight,
• end-to-end measurement,
• non-intrusive manner

s
2 sec
a
7 sec
b
5 sec
c
t
6
Distributed Binning

• Goal
• Have a set of nodes independently partition
  themselves into disjoint bins
• Nodes within a single bin are relatively closer
  to one another than to nodes not in their bin
• Scheme
• A well-known set of machines that act as
  landmarks on the Internet
• Form a distributed binning of nodes based-on
  their relative distances
• A node measures round-trip-time (RTT) to each
  landmark and orders landmarks in order of
  increasing RTT
• Every node has an associated ordering of
  landmarks(or bin)

7
Distributed Binning

• Scheme (Cont.)
• After finding ordering, we calculate absolute
  values of each RTT in ordering as follows
• We divide the range of possible latency values
  into a number of levels.
• Convert RTT values into level number and obtain a
  level vector
• Example
• Level 0? 0-100 ms
• Level 1? 100-200 ms
• Level 2? gt 200ms
• Node As bin becomes l2l3l10 1 2
• Topologically close nodes likely to have same
  ordering and belong to same bin

l2
l1
57 ms
l3
232 ms
A
117 ms
8
Distributed Binning
Distributed Binning Scheme
9
Performance of Distributed Binning

• Even though it is clearly scalable, does it do a
  reasonable job?
• Metric used
• average inter-bin latency average
  latency from a given node to all nodes not in its
  bin
• average intra-bin latency average
  latency from a given node to all nodes in its bin
• A higher gain ratio indicates a higger reduction
  in latency, hence more desirable

10
Performance of Distributed Binning

• Datasets or test topologies
• TS-10K and TS-1K
• Transit-Stub topologies with 10000
• and 1000 nodes respectively.
• 2-level hierarchy
• PLRG1 and PLRG2
• Power-Law Random graph with 1166 and 1779 nodes
• Edge latencies assigned randomly
• NLANR
• Distributed network of over 100 active monitors
• Systematically perform scheduled measurement
  between each other

11
Performance of Distributed Binning

• Other binning algorithms used in experiments
• Random Binning
• Each nodes selects a bin at random
• acts as a lower bound for the gain ratio
• Nearest Neighbor clustering
• Each node is initially assigned to a cluster
  itself.
• At each iteration, two closest clusters are
  merged into a single cluster.
• The algorithm terminated when the required number
  of clusters is obtained
• _

12
Performance of Distributed Binning

• Experiments

Effect of number of landmarks (level1)
Effect of number of levels (landmarks12)
13
Performance of Distributed Binning

• Experiments

Comparison of different binning
techniques(levels1)
14
Topologically-aware construction of overlay
networks

• Two types of overlay networks
• Structured
• Nodes are interconnected in some well-defined
  manner(Application-level)
• Unstructured
• Much less structured like Gnutella,Freenet
• Metric for evaluation

15
Topologically-sensitive CAN construction

• Content-Addressable Network
• Scalable indexing system for large-scale
  decentralized storage applications on the
  Internet
• Built around a virtual multi-dimensional
  Cartesian coordinate space
• Entire coordinate space is dynamically
  partitioned among all the peers, i.e. every peer
  possesses its individual, distinct zone within
  the overall space
• A CAN peer maintains a routing table that holds
  the IP address and virtual coordinate zone of
  each of its neighbor coordinates

16
2D CAN Example
State of the system at time t
Peer
Resource
Zone
x
In this 2 dimensional space, a key is mapped to a
point (x,y)
17
Routing in CAN
y

• d-dimensional space with n zones
• Routing path of length
• Algorithm
• Choose the neighbor nearest to the destination

Peer
(x,y)
Q(x,y)
Query/ Resource
18
Contribution to CAN

• Construct CAN topologies that are congruent with
  underlying IP topology
• Scheme
• With m landmarks, m! such ordering is possible
• For example, if m2, then possible orderings are
  ab and ba
• We partion the coordinate space into m! equal
  sized portions, each corresponding to a single
  ordering
• Divide the space along first dimension into m
  portions
• Each portion is then sub-divided along the second
  dimension into m-1 portions
• Each of these are divided into m-2 portion and so
  on
• When a node joins CAN at a random point, the node
  determines its associated bin based-on delay
  measurement
• According to its landmark ordering, it takes
  place in the correspanding portion of CAN

19
Gain in CAN using Distributed Binning
Stretch for a 2D CAN topology TS-1Klevels1
Stretch for a 2D CAN topology PLRG2levels1
20
Topologically-aware construction of unstructured
overlays

• Aims much less structured overlay such as
  Gnutella, Freenet
• Focusing on the following general problem in
  unstructured overlays
• Optimal overlay is NP-hard, so used some
  heuristic called Short-Long

Given a set of n nodes on the Internet, have
each node picks any k neighbor nodes from this
set so that the average routing latency on the
resultant overlay is low
21
Topologically-aware construction of unstructured
overlays

• Short-Long Heuristic
• A node picks its k neighbors by picking k/2 nodes
  closest to itself and then picks another k/2
  nodes at random
• Well-connected pocket of closest nodes and
  inter-connections to far pockets with random
  picks
• BinShort-Long (Contribution)
• A node picks k/2 neighbors at random from its bin
  and picks remaining k/2 at random

Current Node
Nearby Nodes
Distant Nodes
Other Nodes
22
Gain in Unstructured Overlay using Distributed
Binning
Unstructured overlays TS-10Klevels1landmarks
12
23
Topology-aware server selection

• Replication of content over Internet gives rise
  to the problem of server selection
• Parameter Server load and distance(in term of
  Network Latency)
• _

Replication Server
Client
24
Topology-aware server selection

• Server selection process with distributed binning
  works as follows
• If there exist one or more servers within same
  bin as client, then client is redirected to a
  random server from its own bin
• If no server exists within same bin as client,
  then an existing server whose bin is most similar
  to clients bin is selected at random
• Compared performance to 3 schemes
• Random Client selects server at random
• Hotz Metric Uses RTT measure from a node to well
  known landmarks to estimate internode distance
  (Triangle inequality)
• Cartesian Distance Calculates Euclidean distance
  using level vector of node and selects the server
  with minimum distance
• Measurement for evaluation

25
Topology-aware server selection

• Comparison of different schemes under following
  conditions
• 12 landmarks and 3 levels
• 1000 servers for TS-10K, 100 servers for TS-1K,
  PLRG1 and
• PLRG2 and 10 for NLANR

26
Topology-aware server selection-Node Perspective
CDF of latency stretch for NLANR data
CDF of latency stretch for TS-10K data
27
Conclusion

• Described a simple,scalable,binning scheme that
  can be used to infer network proximity
  information
• Nature of the underlying network topology affects
  behavior of the scheme
• It is applied to the problem of
  topologically-aware overlay construction and
  server selection domains
• Three applications of distributed binning is
  given
• Structured Overlay
• Unstructured Overlay
• Server selection
• A small number of landmarks yields significant
  improvements.
• Can be referred as network-level GPS system
• _

28
Happy end! Thank you for your patience!