routing without flow control hot-potato routing simulation
DESCRIPTION
Project:. Routing Without Flow Control Hot-Potato Routing Simulation. by Larry Bush. Motivation Paper Summary. Overview. Concepts How Algorithm Works Demonstration Results. Algorithm Concepts. Flow Control Hot Potato Routing Greedy Dynamic. Topology. N*N Torus Network. - PowerPoint PPT PresentationTRANSCRIPT
1
Routing Without Flow ControlRouting Without Flow Control
Hot-Potato RoutingHot-Potato RoutingSimulationSimulation
by Larry Bush
Project:
2
MotivationMotivation
Paper Paper SummarySummary
3
OverviewOverview
Concepts
How Algorithm Works
Demonstration
Results
4
Algorithm ConceptsAlgorithm Concepts
Flow Control
Hot Potato Routing
Greedy
Dynamic
5
TopologyTopology
N*N Torus Network
6
7
8
The AlgorithmThe Algorithm
Good-Link / Bad-Link
Home-Run Path
4 priority states: Sleeping … Lowest Priority Active Excited Running … Highest Priority
9
State Change RulesState Change Rules
SleepingActiveExcitedRunning
10
Active PacketActive Packet
DemoDemo
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Running PacketRunning Packet
DemoDemo
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
ConflictsConflicts
Both PacketsConflicts
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
DemoDemo
Both PacketsPlus Other Sleeping Packets
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
GuaranteesGuarantees
O(N) Delivery Time
O(N) Injection Time
106
Packet Delivery Time
0
20
40
60
80
100
120
140
160
Network Diameter (N)
Tim
e S
teps 25% Injecting Routers
50% Injecting Routers
75% Injecting Routers
100% Injecting Routers
107
Average Wait to Inject a Packet
0
5
10
15
20
25
30
35
40
45
50
Number of Rows (N)
Tim
e S
teps
To
Inje
ct
25% Injecting Routers
50% Injection Routers
75% Injection Routers
100% Injection Routers
108
Parallel Speed-Up ( LPs vs. Event Rate )
0
200000
400000
600000
800000
1000000
1200000
1400000
1600000
0 50 100 150 200 250 300
Network Diameter ( N )
Eve
nts
per
Sec
ond
1 Processor
2 Processor
4 Processor
109
Efficiency ( Speed-Up / #PE )
0
0.2
0.4
0.6
0.8
1
1.2
0 50 100 150 200 250 300
Network Diameter ( LPs )
Eff
icie
ncy
( S
peed
-Up
/ PE
)
1 Processor
2 Processor
4 Processor
110
ConclusionConclusion
Algorithm behaved as expected.
ROSS Speed-up is Good.
111
Take AwaysTake Aways
Difficult to analyze theoretically.
Simulation is a good tool for this type of problem.
ROSS is a good tool for this type of simulation.
112
EndEnd
113
Questions/FactsQuestions/Facts
114
Theoretical Model Theoretical Model
vs. vs.
Practical ModelPractical Model
Theoretical– Does not Absorb Sleeping Packets
Practical– Absorbs Sleeping Packets
115
Outline (notes)Outline (notes) Introduction to the problem
– Project scope.– Motivation.
Utilized all free links.– Paper Overview.– Define concepts in model.– Define model.
Destinations chosen randomly– Demonstrate model.
Results– Model results.
Explain Graphs– Delivery Time (Theoretical Model / Practical Model)– Injection Time (Theoretical Model / Practical Model)
Summarize Results
– ROSS performance Speed-up (Parallel vs. Sequential)
– Event Rate – Execution Time
Parameter Effects– Remote Events Sent / Total Events Processed– Roll-backs / Net Events
116
Graphs / ResultsGraphs / Results Results
– Model results. Explain Graphs
• Delivery Time
• Injection Time
– ROSS performance Speed-up (Parallel vs. Sequential)
• Event Rate
117
Project ScopeProject Scope - Sim & Analyzed - Sim & Analyzed - Algo. In Costas’ Paper - Algo. In Costas’ Paper
Paper Overview Paper Overview - New version of HPR. - New version of HPR. - Proved Theoretically. - Proved Theoretically. - Dynamic vs. Static - Dynamic vs. Static
MotivationMotivation- Utilized all free links.- Utilized all free links.- Difficult to analyze theoretically.- Difficult to analyze theoretically.Simulation is a better tool for this typeSimulation is a better tool for this type
of problem.of problem.
118
Network ConceptsNetwork Concepts
Synchronous
Buffer-less
Model of Optical
Switching Network