Self-Organizing Logical-Clustering Topology for Managing Distributed Context Information

Self-Organizing Logical-Clustering
Topology for Managing Distributed
Context Information
M.Sc. Hasibur Rahman
Department of Computer and Systems Sciences,
Stockholm University
2015-09-24
2015/09/24/Hasibur Rahman, Licentiate Seminar, DSV

Why manage Context Information? (1)
• Everyday Things get connected
• Connected Things > human. Today, the ratio is 1.5:1 [Cisco]
• Hundreds of billions by 2020 [Ericsson, Cisco]
• Not the scale only but also expanding in its scope
• Vast network of Context Information (CI) will remain
underutilized if not properly managed
• According to IBM, 90% of CI in IoT is never utilized or acted
upon [IBM 15]

Why manage Context Information? (2)

4 of 21
But how?
• This necessitates proper management of context
information
• Clustering efficient management context
information
• This thesis proposes a distributed clustering
approach namely logical-clustering.
• The idea is to cluster entities (e.g. things in IoT
landscape) based on similar context as opposed to
physical location-based clustering of entities

5 of 21
Problem statement
• An architecture that reflects the real world
scenario
• A model which enables scalable and dynamic
dissemination of context information
• The rapidly changing dynamic environment further
mandates to bring about self-organization
• Performance evaluation

6 of 21
Research question
• How can a self-organizing logical-
clustering topology be realized towards
efficiently managing a vast network of
distributed context information?

7 of 21
Related work
• e-SENSE [Lombriser et al. 07]
• WSN logical neighborhoods [Mottola and Picco 06]
• Architectures for connecting entities and provisioning
context information [JW14]
• Research until now did not focus on real-time distributed
context information management [Miorandi et al. 12]
• A move towards future IoT requires distributed support
such as layering DHT i.e. H-DHT [Zoel et al. 08]
• Controlling heterogeneous entities and networks such as in
IoT exposes many challenges that can be resolved via SDN
[Pedro and Antonio 14]

8 of 21
Contributions
• Demonstarted the challenges that require to enable
distributed CI clustering
• Four research sub-questions (SQs)
– A system architecture [Publication # I]
– Distributed PubSub Model [Publication # III & IV]
– Self-organization [Publication # V]
– Network performance [Publication # II]

9 of 21
System Model (tiered H-DHT)

Example

Communication

13 of 21
Distributed PubSub Model
• To fulfill the requirements for scalable heterogeneous
context information dissemination and logical-sink
synchronization
• By extending the DCXP protocol and evaluated on a
proven IoT platform namely MediaSense
• High PubSub messages/sec
• Fast context-ID matching

14 of 21
Self-organization
• In line with the vision of autonomic computing
and the control loop of IoE, three self-* aspects
have been designed and developed
• DCXP was further extended and two new
primitive functions were added
• The algorithms now enable each entity to
organize itself and further enabled stability and
resilience within logical-clustering
• Resulted in fast discovery of entity and high
discovery accuracy

Autonomic computing

16 of 21
Network Performance
• Mean delay and packet loss ratio were examined
• For variant information flow rate, and different cluster size
and entity per cluster
• 13% mean delay fluctuated when flow rate and cluster size
are increased by 60% and 200% respectively
• PLR increases by 20% for 100% increase in cluster size
each time
• Flow rate increases by about 338 % when packet size is
halved and PLR decreases by 15%

17 of 21
Conclusions
• The thesis has grounded the vision of logical-clustering
• Several artefacts have been designed, developed, and
evaluated
• A fast, scalable and dynamic architecture
• Promising results:
– fast subscription matching
– high PubSub messages and formulas for PubSub outcome
prediction
– fast discovery of entities
– high discovery accuracy
• A template for enabling autonomic computing in IoT has
been portrayed

18 of 21
Future work
• Context-based clustering algorithm needs to be
designed and developed
• Policies to enable autonomic computing in IoT needs to
be explored
• A more concrete and capable sink in IoT such as SDN
controller should be incorporated to see a fully
functional autonomic IoT
• A naming scheme along with UCI can be explored to
ensure unique identification of entity in IoT landscape
• Exploring load-balancing and scheduling algorithms for
sinks can result in higher discovery accuracy
• Heterogeneous interoperability of the sinks should also
be examined

Thanks for your kind attention!
Questions?

20 of 21
References
• [online] Big data and analysis, Internet of Things by IBM link:
http://www-01.ibm.com/software/info/internet-of-things/iot-prod/iot-announcement.html
• J. Walters, “Distributed Immersive Participation – Realising Multi-Criteria
ContextCentric Relationships on an Internet of Things”, PhD thesis, Department of
Computer and Systems Sciences, Stockholm University, November 2014
• C. Lombriser, M. Marin-Perianu, R. Marin-Perianu, D. Roggen, P. Havinga, G.
Troster, “Organizing Context Information Processing in Dynamic Wireless Sensor
Networks”, Proc. ISSNIP,pp. 67-72,December 2007
• L. Mottola and G. P. Picco: Logical Neighborhoods: A Programming Abstraction for
Wireless Sensor Networks, Distributed Computing in Sensor Systems, Volume
4026, 2006, pp 150-168
• H. Gellersen, A. Schmidt, M. Beigl, "Multi-Sensor Context-Awareness in Mobile
Devices and Smart Artifacts", Mobile Networks and Applications, vol. 7, no.5,
pp.341-351, 2002

21 of 21
References (2)
• Pedro Martinez-Julia, Antonio F. Skarmeta, white paper ” Extending the
Internet of Things to IPv6 with Software Defined Networking”
• T. Kanter et al.., “MediaSense – an Internet of Things Platform for
Scalable and Decentralized Context Sharing and Control,” In: ICDT
2012,, The Seventh International Conference on Digital
Telecommunications, pp. 27-32, April 2012
• S. Zoels, Z. Despotovic, W. Kellerer:, “On hierarchical DHT systems- An
analytical approach for optimal design”, Computer Communications,
Volume 31(3), Page 576-590, 2008
• R.A. Baloch and Noel Crespi. Addressing context dependency using
profile context in overlay networks. In Consumer Communications and
Networking Conference (CCNC), 2010 7th IEEE, pages 1–5. IEEE,
January 2010

Self-Organizing Logical-Clustering Topology for Managing Distributed Context Information

More Related Content

What's hot (17)

Viewers also liked (10)

Similar to Self-Organizing Logical-Clustering Topology for Managing Distributed Context Information (20)

Recently uploaded (20)

Self-Organizing Logical-Clustering Topology for Managing Distributed Context Information