Using Collaborative Agents to Enrich Service Environments

1

• Using Collaborative Agents to Enrich Service
  Environments
• Olga Ratsimor
• Balaji Kodeswaran

2
Problem Statement

• Wide disparity between the capabilities of wired
  and wireless networks
• The wireless devices face frequent and possibly
  prolonged disconnections and bandwidth is limited
  
• Variation in capabilities of mobile devices
• Laptop vs. iPAQ vs. Palm vs. Cell phones
• Wireless devices are resource limited in terms of
  processing power, battery etc.
• Proliferation of wireless services and increased
  sophistication pushes the limits of wireless
  devices
• Traditional text based news services have been
  enhanced to offer graphical and audio-visual
  multimedia content.

3
Problem Statement (cont.)

• In ad hoc wireless networks, devices communicate
  with each other (within constrained boundaries)
  to use/provide services. There is no external
  coordination to improve overall service
  availability
• Infrastructure wireless networks enforce a client
  server model between the mobile user and the
  base stations. This model is too restrictive and
  requires base stations to be strategically placed
  so that services can be offered to mobile clients

4
Proposed Solution

• MH capabilities used to intelligently compose
  services that are best suited for that MH
• Content/Data used by the services must be
  intelligently packaged and strategically
  distributed to maximize efficiency of the overall
  system
• Use profiles/heuristics to proactively
  inject/distribute services into an ad hoc
  environment so as to improve the statistical
  probability of service availability

5
List of Components

• Service Portals
• Base stations that host services and are
  connected to a wireline network
• Mobile Devices
• Agent Platform
• Services
• Service Specification
• Service Agents
• Service Data Volumes

6
Network Model

• The network is comprised of two distinct types of
  zones
• Landing Zones
• Mobile Devices in this zone can communicate with
  a Service Portals (infrastructure)
• Transit Zones
• MH in this zone can communicate with peers only
  (ad hoc)
• Combination of infrastructure and ad hoc wireless
  network concepts
• A Mobile Device can talk to other devices in its
  environment
• A Mobile Device can also talk to fixed wireline
  components like mobile support stations or base
  stations

7
Service Portal

• Service Portals act as base stations and are
  located through out the network
• Each Portal is aware if all of its immediate
  neighbors
• Portals perform following duties
• Actively advertise presence and host different
  types of services
• Perform dynamic data/content management for the
  different services so that MHs are offered only
  data that they can handle/use
• Dynamically create Service Data Volumes that
  are distributed to MHs so that an MH is not
  required to download all data needed for a
  Service at once
• Caching communicate with neighboring portals to
  inform them of possible future service demands in
  their vicinity
• Monitor the usage patterns of services on a MH
  passing through a Landing Zone to extrapolate
  what services/content may be required in
  neighboring Transit Zones and schedules to have
  these services/content delivered through other
  MHs that are heading towards these zones

8
Service Specification

• Description of the Service
• Expressed using descriptive languages
• Expresses high level requirements for a service
• News paper reader requires a UI display
• Audio player requires speakers
• Audio recorder requires a microphone etc.

9
Service Agent

• Each Service specification is associated with
  multiple implementations called Service Agents
  that implement that specification
• Service Agents can be migrated to a MH on demand
• Service Agents can also be automatically dropped
  from a MH when no longer needed
• Service Agents are provided with Service Data
  Volumes on which they operate

10
Service Data Volume

• Service data is pre-divided into Data Units. Data
  Unit is the smallest unit of data
• Articles or Individual pages of a News paper
• Each Song of musical score
• Each Data Unit could be of varying size. Size
  here depends on the service specification
• Words on a Page for a news reader
• Minutes for a song
• Multiple Data Units are aggregated into Service
  Data Volumes for distribution to MHs

11
Mobile Host

• Wireless devices with varying capabilities
  running a thin Agent Platform
• Determine if the vicinity is a Landing Zone or a
  Transit Zone
• Communicate with peers and with Service Portals
• Provide APIs that allow for device capability
  discovery
• Support of dynamic loading and unloading of
  Service Agents and Service Data Volumes
• Profile Service Agents
• Currently registered Service Agents, running
  times, etc
• User actions are logged by respective Service
  Agents
• Which pages of a newspaper has the user read

12
Surveyor

• At start up the Surveyor Agent jumps into the
  device and evaluates device capabilities
• Surveyor composes a Capability Report which is
  sent to the Service Portal
• Depending on this Capability Report the Service
  Portal sends a list of appropriate services to
  the device
• User selects desired service(s)

13
The Surveyor
What can your Device handle?
User selects services
14
NUMI Flavors

• Service Distribution Modes
• On Demand
• Relies on logs on mobile hosts
• Proactive
• States are maintained at Service Portals that
  track expected user mobility. Service Portals use
  this to pre-equip environments.
• MH mobility characteristics
• Direction aware
• Caching is optimized
• Direction unknown
• Conservative caching is used (all neighboring
  portals cache)

15
On Demand Service Distribution

• The device receives the appropriate Service
  Agent(s)
• The device receives the Service Data Volume,
  enough to last until the next Service Portal (the
  longest hop)
• If the direction of the device movement is not
  known the Portal notifies all its neighboring
  Portals about services that have been recently
  requested
• The neighboring Portals preload the expected
  Service Data Units
• The compilation of Volumes does not happen till
  the MH arrives at that Landing Zone
• When the Mobile Host arrives it receives the next
  Service Data Volume

16
Proactive Service Distribution

• In addition to on demand service distribution
• Neighboring Portals are notified of expected
  time of arrival
• This state is used to proactively distribute
  services if the MH does not arrive on time
• If the direction of the MH movement is known then
  only the next hop Portal is notified. Otherwise
  all its neighboring Portals are notified

17
Service Distribution
5 min
15 min
3 min
18
On Time Mobile Host Arrival
Time t15
Time t 10
15 min
Time t 0
19
Rest Stop Scenario

• The Mobile Device could stop along the way.
• When MH is about to run out of service data it
  starts looking for the next Service Data Volume.
• Service Data is available in the neighborhood
• Neighborhood provides the requested data
• Service Data is unavailable
• passing Mobile Hosts log requests
• Portals monitor the logs of incoming MHs
• Portals identifies missing services and and
  arranges to deliver the services to the
  neighborhood
• In addition a Portal can inform its neighboring
  Portals about missing the Services and Data.

20
Request for Service Continuation
Time t15
The High Volume Traffic with Rest Stop (On
Demand)
21
Request for Service Continuation
Time t15
The low traffic with rest stop and inter portal
communication (On Demand)
22
Proactive Service Transfer

• The Mobile Host might not be resource rich. It
  could be unable to store enough information till
  the next Portal
• If the direction is known the current Portal can
  tell the next hop Portal that the next hop Portal
  should send the the next chunk of data with some
  other Mobile Host that is heading towards the
  Mobile Host in need.

23
Proactive Service Transfer
24
Group Travel

• Mobile Device requires service, however does not
  have enough capacity to store the minimal Service
  Data Volume
• If there is a group of Mobile Hosts that are
  traveling along the same route the Service Data
  can be shared among the devices
• If route is not known the following heuristic can
  be used
• The statistical probability of Service Data
  Volume use should be proportional to the number
  of hosts it is distributed to

25
Multi-Hop Known Route

• Extension of our Proactive Service Distribution
  scheme
• Look beyond next hop neighboring Portals
• Complete route of device used to inform all
  Portals on route about the service needs of this
  device and the expected times of these needs
• Portals repeatedly update other Portals on the
  route when they detect changes in mobility
  characteristics of the device and service usage
  patterns

26
20 min
10 min
Proactive Service Transfer With multi Hop Route
Time t5
Time t10
27
20 min
10 min
Proactive Service Transfer With multi Hop Route
Time t5
Time t10
Time t15
Time t20
28

• Thank You!
• Q.E.D.