Working Group on Applications

1
Working Group on Applications

• H. Li
• J.O. Pereira
• R. Baldoni
• E. Riviere

2
Gossip Based Applications

• Good
• Adaptation
• Robustness
• Self-organization
• Large scale
• Unknown Environment
• ..

• Bad
• Efficiency
• Determinism
• Lack of precise and comparable specifications
• ..

3
Application Layer
Semantic setting
Semantic properties (e.g. clustering through
common interests, semantic peersampling etc)
Semantic Gossip Instantiation Aggregation level
and communication layer
LS
LS
LS
LS
LS
LS
LS
LS
Global properties (e.g. connectivity,
peersampling)
Physical Gossip Instantiation Aggregation level
and communication layer
LS
LS
LS
LS
LS
LS
LS
LS
Physical Setting
network
4
Position

• Many systems employing gossiping use an ad-hoc
  design. This limits the number of applications
  that could exploit the benefits of gossiping. We
  avocate the need for a design framework at the
  conceptual level based on generic building
  blocks.

5

• The design framework is basically split into two
  main layers the semantic one and the physical
  one.
• The goal of the physical layer is to provide at
  the upper layer an abstraction including services
  for disseminating information and peer sampling.
  Each node include a local state which is used to
  guarantee global connectivity as well as previous
  services.

Peer-sampling
Dissemination
Overlay management protocol
TCP/IP networks
6

• The goal of the semantic layer is to build the
  logical network compliant to the application
  needs. In this sense it provides an abstraction
  to applications including services for
• group composition,
• logical failure detection,
• group dissemination,
• semantic peer-sampling
• Aggregation
• Eventual consistency
• Example group composition based on
• Properties of the local state (Monica Belluccis
  fans club)
• Properties of a subset of local states
  (geographycal proximity)
• Global property (k groups of some size h)
• Properties instrumented by Applications
  (integrated approach)

7
Applications

• Applications

File sharing
Streaming
Pub/sub
Games
Sensors networks
System Management
File system synch
Virus Containment
Logical failure detection
Semantic peer-sampling
Group Composition
Dissemination
Semantic Local State
Semantic Layer
gossiping based logic for semantic overlay
Peer-sampling
Dissemination
Physical layer
Local State
Overlay management protocol
TCP/IP networks