Mobius: a SociallyAware Distributed Systems Infrastructure

1
Mobius a Socially-Aware Distributed Systems
Infrastructure

• OR
• Whats this project about and
• how you could contribute

Anda Iamnitchi anda_at_cse.usf.edu
2
Background

• Large-scale distributed systems
• P2P systems, Grids, PlanetLab
• Resource management and resource sharing
• Data management where to place data, how many
  copies, how to transfer it fast to the user
  (BitTorrent? Parallel transfers? Etc)
• Storage management
• Scheduling
• Traditional approach improve performance based
  on systems metrics
• Size of data, network capacity, workload
  characteristics, etc.

3
Mobius Main Research Question
Jane
Bob
Alice

• Can we exploit social knowledge in the design of
  distributed systems?
• What social knowledge? How to collect it?
• What components of a distributed systems can
  benefit from social knowledge?
• How to deal with privacy issues?

4
The Step to Mobile Computing Geo-Social
Knowledge and Context

• Social knowledge best collected from mobile
  users
• Powerful devices, largely deployed
• State of the art
• Big Brother centralized solutions
• Rudimentary social applications (restricted to
  svc. provider)
• Facebook, MySpace, etc. ported onto mobile phones
• Matching apps. -I like dark chocolate - Oh,
  me too! Lets meet!

5
Mobius

• Decentralized 2-tier architecture
• Users put together their PCs and mobile devices
  to create a community infrastructure
• P2P tier provides support to mobile applications
• Persistent services (user-deployed or service
  provider)
• Social state management
• Adaptive to the geo-social context to
• Enable energy efficiency
• Provide access to applications not supported on
  mobiles
• Improve performance
• Mobile tier
• Runs mobile applications and collects geo-social
  information using ad hoc communication

5
6
Application Scenario Community Multimedia
Sharing (1)
Mike
Alice, Mike Jane friends
Jane
Alice
Download mobile application for Bobs service
Mobile tier
P2P tier
Register service

• Bobs service enable mobile users to upload
  share multimedia content
• Sharing community is specified according to type
  and strength of social ties

Bobs service
Service discovery service
7
Application Scenario Community Multimedia
Sharing (2)
Jane
Alice
Mike
Mobile tier
P2P tier
Event notification service
Bobs service
Service discovery service
Janes PC
8
P2P Tier Architecture
9
Mobile Node Architecture
9
10
Other Application Scenarios

• Identify important meetings and silence all
  incoming calls except emergencies (from a white
  list)
• Geo-social recommendations
• Evacuation route to end up with friends and
  family?
• Recommendations about places?
• Limit worm propagation based on a combination of
• Trust inferred from social information
• Support from the P2P layer (via computation,
  information gathering, services, etc).

11
Collaboration with NJIT

• Our main responsibility research,
  implementation, testbed deployment of
• The P2P layer
• Privacy/security issues
• NJITs main responsibility the mobile layer
• SmartCampus testbed 200 mobile phones
• Profs. Cristian Borcea and Quentin Jones and
  their students
• Collaboration administravia
• Conf calls every 2 weeks _at_ 1030am (when
  productive)
• First call Friday, September 26, 1030am (skype)
• Face-to-face meetings later (if productive)
• Most interactions blog and email group (created)

12
(Some) Research Components

• Collect social knowledge
• What is social knowledge? What is social context?
• Needs to be relevant to our objectives
• Can we combine social knowledge from
  Facebook-like sites with info gathered from
  mobile phones?
• Problem with Facebook rule.
• What are the privacy concerns? (bits of
  information vs. aggregated information)
• Define the API between the P2P and mobile layers
• Identify all basic services
• Start with the target applications. Are there
  other applications that must be considered now?

13
Research Components (cont)

• Socially-aware P2P Overlay construction
• Decide neighbors based on geo-social knowledge?
• Metric for privacy? Trust? Incentives? (how do we
  measure benefits?)
• Socially-aware Data Management
• Choose data location based on social ties? (trust
  incentives. Can privacy benefit?)
• Distinguish between state data and content data?
• Version for potential MS thesis
• Socially-aware Service Migration/Instantiation
• When and where to migrate/replicate a service in
  the P2P infrastructure?
• When and where to offload an application from a
  mobile phone to a peer?

14
Your ideas?