Download Document

Optimaalisen tietoliikenneverkon
suunnittelu
Jukka K. Nurminen
Nokia Research Center
[email protected]
1
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Cellular Network Planning
1. Radio network
2. Transmission network
3. Packet network
2
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Radio Network Planning
• Composite Coverage displays
• Signal propagation models
• Heuristic frequency allocation algorithms
3
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Transmission Network Planning Tool
Plan the transmission network to
connect GSM basestations with
• minimum cost
• high reliability
• easy maintenance
• flexibility for growth
•…
• Shortest path routing
• Two disjoint paths routing
• Multilayer routing
• Topology generation
•Min spanning tree
•Traveling salesman
4
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Packet Network Modeling
• IP, ATM
• Parameters (random variables):
• packet inter-arrival time
• number of packets in a packet call
• packet length
• reading time between packet calls
Packet call
time
Packet
arrivals
5
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Reading time
between packet
calls
Interrupted Poisson Process (IPP)
ON period
OFF
OFF
a
-exponentially distributed with
mean 1/ON
OFF period -exponentially distributed with
mean 1/OFF
ON
Arrival rate -exponentially distributed with
parameter a
ON
Packet length
ON
period
6
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
OFF
period
Queuing Models
Queuing models are suitable for describing (modeling) packet switched
nodes and networks
•  - packet arrival rate
• L – average packet length
• C – link capacity
•1/m – average service time
• W – waiting time in the
buffer
• d – delay
• N – number of packets in
the system
• r - traffic load
m
Arriving
packets
, L
buffer
Departin
g packets
server
C
Outgoin
g
link
The mean delay in the node is equal to the sum of the mean waiting time in
the buffer and the mean service time
E[d]=E[W]+1/ m = E[W]+L/C
7
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Peer-to-peer Network Analysis
1. Content sharing
2. Distributed computing
8
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Content sharing
MP3 music, pictures, DVD movies, etc.
• Napster
Centralized
• Systems based on
FastTrack
KaZaA
Grokster
Morpheus (earlier)
• Gnutella
LimeWire
Shareaza
Morpheus (newer)
Distributed
9
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
How it works?
Napster
Gnutella
2. Propagate search
5. Transfer
4. Hit found
3. Transfer
1. Search
1. Register
Search A.mp3
10
© NOKIA
2. Search
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
0. Connect
Topologies for Content Sharing
• Reported: power law (“fit get fitter”, “rich get richer”) evolved into multimodal
very dynamic
cannot be easily emulated
• Chosen options:
• Random: nodes have random number of connections
• Semi: nodes have random number of connections, at least 1
• ConMesh: nodes connect to existing network
• ConStars: ConMesh + leaf nodes
Random mesh
11
© NOKIA
Semi- random mesh
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Connected mesh
Connected star
Simulation results
12
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Distributed computing
• Currently on distributed
PCs
• Can such applications run
on mobile phones?
13
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
+
mobile
phone
Root server
Client/Server 1
Client 2
Client 3
Client 4
Client 5
Personal Java on Nokia Communicator
14
© NOKIA
Presentation_Name.PPT / 13-11-2003 / Jukka K. Nurminen
Lucas-Lehmer test:
2p-1 is prime if and only if S(p-1) = 0 (mod 2p-1),
where S(1) = 4 and S(k+1) = S(k)2-2.