2 October 2006 - PhD Viva Presentation

1. An Efficient Reactive Model for
Resource Discovery in DHT-Based
Peer-to-Peer Networks
James Salter
Supervisor: Dr Nick Antonopoulos
2 October 2006

2. Outline
 Introduction and Background
 ROME Architecture
 Node Processes
 Evaluation
 Conclusions and Future Work

3. Client/Server Networks
 Clients send requests to/via a central server
 Small #messages
 Single point of failure

4. Peer-to-Peer Networks
 No central server
 Node to node connections
 Resilient
 Large #messages
File sharing, distributed computing, instant messaging

5. Napster

6. Gnutella

7. Chord
 Structured Peer-to-Peer architecture
 Well-known in the research field
 Combine advantages of Napster and
Gnutella-like architectures:
 An index of resources (Napster)
 distributed over multiple nodes (Gnutella)
 Based on Distributed Hash Tables
 Simple lookup mechanism
 Given a key, it will return associated value(s)

8. 10, 16, 28, 34
Distributed Hash Tables
 Hash functions map data keys to buckets
 Keys are stored in buckets in index table
0
1
2
3
4
5
f(x) = x mod 6
f(15) = 15 mod 6 = 3
12, 24, 30
19, 61
20, 26, 56
3, 9
23, 35, 65
f(36) = 36 mod 6 = 0
Distributed Hash Tables
 Each node hosts part of the index (one or
more buckets)

9. Chord
20
0 1
8
12
15
2 3 4 5 6 7 8

10. Chord
log2(n) hops worst case
½log2(n) hops average

11. ROME Concept
 Message cost is proportional to number
of nodes in network (n)
 Reduce n, reduce message cost
 Goal: Keep the ring “just big enough”
 Must always support current workload
 Not unnecessarily large
 Workload should determine ring size,
not number of nodes in the network
 Adding functionality to Chord

12. ROME Architecture
ROME
Chord
Lower
Layers
Traffic
Analyser
Node
Actions
ROME Data

13. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

14. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

15. Node Workload Monitoring
Zero LimitLower
Threshold
Target Upper
Threshold

16. Node Workload Monitoring
Zero LimitLower
Threshold
Target Upper
Threshold
Underloaded

17. Node Workload Monitoring
Zero LimitLower
Threshold
Target Upper
Threshold
Underloaded Normal

18. Node Workload Monitoring
Zero LimitLower
Threshold
Target Upper
Threshold
Underloaded Normal Overloaded

19. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

20. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

21. Replace Action
Overloaded
v0.2 5 cur_wl, targetA

22. Replace Action
 Search node pool to find a node with:
NPNode.LowerThreshold < OLNode.Workload AND
NPNode.UpperThreshold > OLNode.Workload
 Break ties by selecting best
quality node
 Percentage of heartbeats
received from node since
initial registration
 Measure of node reliability

23. Replace Action
Overloaded
v0.2 6 replace, node=Bbs
v0.2 7 chordID=21A

24. Replace Action
v0.2 1 tgt, thresholdsA

25. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

26. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

27. Add Action
 No nodes found to replace overloaded
node
 Add a node to share enough
workload to restore overloaded
node to normal workload
NPNode.LowerThreshold <
(OLNode.Workload - OLNode.Target)
AND
NPNode.UpperThreshold >
(OLNode.Workload - OLNode.Target)

28. Add Action
Overloaded
v0.2 6 add, node=C, C.tgt, C.thresholdsbs

29. Add Action
16 17 18 19 2015 21
ID: 21ID: 14
workload
ID: ??

30. Add Action
Overloaded
v0.2 7 chordID=18A

31. Add Action

32. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

33. ROME Node Process
Monitor
Workload
Attempt to
Replace Node
Attempt to Add
New Node
Attempt to
Remove Node
Pause
Overloaded Underloaded
Normal
Success
Failure

34. Remove Action
Underloaded
v0.2 8 current_wlG

35. Remove Action
 If node is removed, successor
becomes responsible for its portion of
keyspace (+ workload)
 Must check successor will not become
overloaded: prevent chain reactions
Succ.UpperThreshold >
Succ.Workload + ULNode.Workload

36. Remove Action
Underloaded
v0.2 9 Accept/DeclineH

37. Remove Action
v0.2 1 tgt, thresholdsA

38. Additional Issues
 Node Locking
 Prevent chain reactions/overcompensation
 Workload from a single key
 Add action not attempted: assumption that
keys are atomic
 Ring collapse vulnerability
 Monitoring interval to stop too many
concurrent changes
 High volume of replaces
 Switch off replace action if few nodes in pool

39. Dynamic Ring Performance
 Reduction in hop count by controlling
network size based on theoretical work:
 Max hops per lookup = log2(n)
 Mean hops per lookup = ½log2(n)
 If ring A < ring B, then log2(A) < log2(B)
 - in a static ring with correct routing information
 Does this hold true in more realistic
dynamic scenarios, with nodes
joining/leaving or failing?

40. Simulation: ROME vs Chord
 Two Chord rings, one running ROME
 1000 available nodes
 Node capacity: 100-400 workload units
 10 node joins per time tick
 10 node failures per tick
 500 lookups per tick
 ROME Thresholds: 5% Lower, 95% Upper
 Target Workload: 50%
 Standard Chord maintenance/update routines
run every clock tick

41. Workload Variation
0
50000
100000
150000
200000
250000
0 2000 4000 6000 8000 10000 12000
time
workload

42. 0
1
2
3
4
5
6
0 2000 4000 6000 8000 10000 12000
time
meanmessagespersuccessfulquery
Messages per Query
ROME Chord

43. Cumulative Message Savings
0
1
2
3
4
5
6
7
8
0 2000 4000 6000 8000 10000 12000
time
lookupmessagesavings(cumulative
millions)

44. Number of Nodes in Each Ring
ROME Chord
0
100
200
300
400
500
600
700
800
900
1000
0 2000 4000 6000 8000 10000 12000
time
nodesinring

45. Query Success Rate
0%
20%
40%
60%
80%
100%
0 2000 4000 6000 8000 10000 12000
time
%successfulqueries
ROME Chord

46. Maintenance Cost
0
1000
2000
3000
4000
5000
6000
7000
8000
0 2000 4000 6000 8000 10000 12000
time
maintenancemessages
ROME Chord

47. Thresholds

48. Thresholds
LT=5%, UT=95%
LT=25%, UT=75%
LT=45%, UT=55%
0%
20%
40%
60%
80%
100%
0 2000 4000 6000 8000 10000 12000
time
%successfulqueries

49. Churn Rate (Failure is Good!)
0
200
400
600
800
1000
0 2000 4000 6000 8000 10000 12000
time
nodesinring
0 fail/tick
1 fail/tick
5 fail/tick
10 fail/tick
20 fail/tick
30 fail/tick

50. Churn Rate (Failure is Good!)
0
1
2
3
4
5
6
7
0 2000 4000 6000 8000 10000 12000
time
meanmessagespersuccessfulquery
0 failures/tick
1 failure/tick
5 failures/tick
10 failures/tick
20 failures/tick
30 failures/tick

51. Churn Rate (Failure is Good!)
0%
20%
40%
60%
80%
100%
0 2000 4000 6000 8000 10000 12000
time
%successfulqueries
0 failures/tick
1 failure/tick
5 failures/tick
10 failures/tick
20 failures/tick
30 failures/tick

52. Limitations of ROME
 Requires workload to be less than node
capacity
 May not be appropriate if likely to be near-equal for
majority of network’s lifetime
 Optimisations occur at the node level
 Always yields a globally optimum solution?
 Based on Chord and DHT architectures
 Any issues found (probably) inherited by ROME
 Increasing workload, failed bootstrap server
 No more nodes will be added – these would be present in
standard Chord ring, so ROME ring would drop more
workload

53. Potential Applications
Similar to Chord applications
 DNS lookup services
 File storage systems
 Simple databases
 Messaging
 Service Discovery
 Distributed Processing
Where workload is likely to be lower than
capacity offered by connected nodes

54. Future Work
 Remove reliance on single server
 Share node pools between multiple bootstrap
servers using G-ROME
 Combinations of actions
 Apply ROME concepts to other P2P
networks
 Use in unstructured networks?
 Applications in other domains
 Wireless/ad-hoc networks with dynamic
machine joins/leaves

55. Conclusions
 Proposed ROME, a layer running on
top of Chord
 Chord routes messages in O(log2 n) hops,
where n is number of nodes in ring
 ROME controls size of underlying Chord ring
 Simple set of actions to add/remove nodes
 Simulations show ROME can reduce
lookup cost vs standard Chord ring
 Platform for further work
 Enhance ROME, use as building block for new
services, apply to other domains

56. Publications List
J Salter and N Antonopoulos, “An Optimised Two-Tier P2P Architecture for Contextualised
Keyword Searches”, Future Generation Computer Systems, 2007.
G Exarchakos, J Salter and N Antonopoulos, “G-ROME: A Semantic Driven Model for Capacity
Sharing Among P2P Networks”, to appear in Internet Research.
G Exarchakos, J Salter and N Antonopoulos, “Semantic Cooperation and Node Sharing Among
P2P Networks”, Sixth International Network Conference (INC 2006).
J Salter and N Antonopoulos, “The CinemaScreen Recommender Agent: A Film Recommender
Combining Collaborative and Content-Based Filtering”, IEEE Intelligent Systems, 2006.
N Antonopoulos, J Salter and R Peel, “A Multi-Ring Method for Efficient Multi-Dimensional Data
Lookup in P2P Networks”, 2005 International Conference on Foundations of Computer
Science (FCS ’05).
J Salter, N Antonopoulos and R Peel, “ROME: Optimising Lookup and Load Balancing in DHT-
based P2P Networks”, 2005 International Conference on Parallel and Distributed
Processing Techniques and Applications (PDPTA ’05).
J Salter and N Antonopoulos, “ROME: Optimising DHT-based Peer-to-Peer Networks”, Fifth
International Network Conference (INC 2005).
N Antonopoulos and J Salter, “Efficient Resource Discovery in Grids and P2P Networks”,
Internet Research, 2004.
J Salter and N Antonopoulos, “An Efficient Fault Tolerant Approach to Resource Discovery in
P2P Networks”, UniS Computing Sciences Report CS-04-02, 2004.
N Antonopoulos and J Salter, “Improving Query Routing Efficiency in Peer-to-Peer Networks”,
UniS Computing Sciences Report CS-04-01, 2004.
J Salter and N Antonopoulos, “An Efficient Mechanism for Adaptive Resource Discovery in
Grids”, Fourth International Network Conference (INC 2004).
N Antonopoulos and J Salter, “Towards an Intelligent Agent Model for Resource Discovery in
Grid Environments”, IADIS International Conference Applied Computing 2004.