14
NGMAST 2008 - Mobile DHT Energy 1 Optimizing Energy Consumption of Mobile Nodes in Heterogeneous Kademlia-based Distributed Hash Tables Imre Kelényi Budapest University of Technology and Economics Department of Automation and Applied Informatics Jukka K. Nurminen Nokia Research Center

Imre Kelényi Budapest University of Technology and Economics

  • Upload
    ulla

  • View
    52

  • Download
    1

Embed Size (px)

DESCRIPTION

Optimizing Energy Consumption of Mobile Nodes in Heterogeneous Kademlia-based Distributed Hash Tables. Imre Kelényi Budapest University of Technology and Economics Department of Automation and Applied Informatics Jukka K. Nurminen Nokia Research Center. Agenda. - PowerPoint PPT Presentation

Citation preview

Page 1: Imre Kelényi Budapest University of Technology and Economics

NGMAST 2008 - Mobile DHT Energy 1

Optimizing Energy Consumption of Mobile Nodes in Heterogeneous Kademlia-based Distributed Hash Tables

Imre KelényiBudapest University of Technology and Economics

Department of Automation and Applied Informatics

Jukka K. NurminenNokia Research Center

Page 2: Imre Kelényi Budapest University of Technology and Economics

NGMAST 2008 - Mobile DHT Energy 2

Agenda Distributed Hash Tables (DHT)

DHT on mobile phones

Probabilistic message dropping

Measurements and results

Analytical model

Conclusion

Page 3: Imre Kelényi Budapest University of Technology and Economics

NGMAST 2008 - Mobile DHT Energy 3

Distributed Hash Tables (DHT) Peer-to-peer systems that enable the storage

and retrieval of <key,value> pairs in a completely distributed and scalable way

Distributed storage systems, file sharing, service discovery, P2P SIP, etc.

DHT nodes communicate with each other via messages (e.g. over UDP datagramms)

There has not been any mobile implementation

Page 4: Imre Kelényi Budapest University of Technology and Economics

NGMAST 2008 - Mobile DHT Energy 4

DHT on mobile phones Mobile Kademlia for Symbian OS

Implements the Kademlia DHT protocol It can connect to the global Mainline

BitTorrent DHT network (usually has more than 1 million concurrent users)

Open source project More information:

http://www.aut.bme.hu/MobileDHT

Page 5: Imre Kelényi Budapest University of Technology and Economics

Problem with the mobile DHT When connected to a large DHT, the energy consumption

is so high that the battery is depleted in a couple of hours

This is mostly the result of the large incoming traffic: requests coming from other DHT nodes

NGMAST 2008 - Mobile DHT Energy 5

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0:00

:00

0:15

:00

0:30

:00

0:45

:00

1:00

:00

1:15

:00

1:30

:00

1:45

:00

2:00

:00

2:15

:00

2:30

:00

2:45

:00

3:00

:00

W

Page 6: Imre Kelényi Budapest University of Technology and Economics

First solution: block all requests It is possible to block all incoming requests and

thus operate in client-only mode

Significant energy conservation but also decreases DHT performance

NGMAST 2008 - Mobile DHT Energy 6

0.000

0.200

0.400

0.600

0.800

1.000

0:00

:00

0:15

:00

0:30

:00

0:45

:00

1:00

:00

1:15

:00

1:30

:00

1:45

:00

2:00

:00

2:15

:00

2:30

:00

2:45

:00

3:00

:00

W

Page 7: Imre Kelényi Budapest University of Technology and Economics

Probabilistic message dropping Concept: do not block all incoming requests but a

fraction of them Energy is conserved The node still participates in the DHT to some degree

Implementation: if an incoming request is received, it is dropped with a certain probability Pdrop

The larger Pdrop is, the more messages are blocked and more energy is conserved

Full peer: Pdrop=0, client-only: Pdrop=1

It does not require any protocol modifications in the other nodes (backward compatible)!

NGMAST 2008 - Mobile DHT Energy 7

Page 8: Imre Kelényi Budapest University of Technology and Economics

Probabilistic message dropping Protocol requirement

The DHT must store multiple replicas of the stored values

Energy is conserved Responses are not sent out

Less requests are received

NGMAST 2008 - Mobile DHT Energy 8

Page 9: Imre Kelényi Budapest University of Technology and Economics

Measurement results Performed with a Nokia N95, connected to

the Mainline BitTorrent DHT

Average of 3-hour-long sessions

NGMAST 2008 - Mobile DHT Energy 9

Page 10: Imre Kelényi Budapest University of Technology and Economics

Message traffic statistics

NGMAST 2008 - Mobile DHT Energy 10

Page 11: Imre Kelényi Budapest University of Technology and Economics

Analytical Model on DHT Performance Increasing Pdrop decreases the performance of

the DHT, but to what extent?

Our analytical model characterizes the lookup latency of the DHT as a function of Pdrop and the ratio of mobile nodes in the network m

NGMAST 2008 - Mobile DHT Energy 11

Page 12: Imre Kelényi Budapest University of Technology and Economics

Analytical Model on DHT Performance Expected lookup latency in Mainline BitTorrent DHT

as a function of m (ratio of mobiles) and Pdrop

NGMAST 2008 - Mobile DHT Energy 12

Page 13: Imre Kelényi Budapest University of Technology and Economics

Conclusion Probabilistic message dropping enables

heterogeneity in DHTs without protocol modification

It enables mobile nodes with scarce resources, most importantly battery capacity, to join DHTs

Finding the optimal value for Pdrop is still an open question Dynamical adjusting

NGMAST 2008 - Mobile DHT Energy 13

Page 14: Imre Kelényi Budapest University of Technology and Economics

Thank You! More information:

http://www.aut.bme.hu/mobiledht

http://amorg.aut.bme.hu

Questions?

NGMAST 2008 - Mobile DHT Energy 14