context aware migratory services aware migratory services
play

Context- -aware Migratory Services aware Migratory Services - PDF document

Context- -aware Migratory Services aware Migratory Services Context in Ad Hoc Networks* in Ad Hoc Networks* Rutgers-Helsinki Ph.D. Student Workshop on Spontaneous Networking 8-12th May 2006 Oriana Riva University of Helsinki, Dep. of


  1. Context- -aware Migratory Services aware Migratory Services Context in Ad Hoc Networks* in Ad Hoc Networks* Rutgers-Helsinki Ph.D. Student Workshop on Spontaneous Networking 8-12th May 2006 Oriana Riva University of Helsinki, Dep. of Computer Science oriana.riva@cs.helsinki.fi *Joint work with Liviu Iftode (Rutgers), Cristian Borcea (NJIT) and Tamer Nadeem (Univ. of Maryland) Connecting to Internet through Connecting to Internet through ad hoc networks ad hoc networks Internet How else can we exploit the How else can we exploit the ad hoc network ad hoc network characteristics to support a characteristics to support a new class of applications? new class of applications? 1

  2. Ad hoc networks as service providers Ad hoc networks as service providers � Enable a new class of services specific to ubiquitous computing environments � A Acquire, process, disseminate real-time � information � Targets: region, activity, entity Traffic Information from region at constant Traffic Information from region at constant distance in front of the driver distance in front of the driver � Requirements: • Service must be aware of its context • Service must adapt to user’s context 2

  3. Entity Tracking Service Entity Tracking Service � Requirements: • Service must be aware of its target • Service must preserve its execution state over time Spontaneous Monitoring Service Spontaneous Monitoring Service � Requirements • Transfer code from node to node • The service must be aware of its context 3

  4. Requirements for services in ad hoc Requirements for services in ad hoc networks networks � Context-awareness • dynamism of services and request targets � User-driven adaptability • dynamism of user needs and operational context � Service continuity • due to context changes a node may become incapable of hosting the service any longer • need to support stateful interaction � On-demand code distribution • nodes do not possess the code for any type of service Outline Outline � Motivations � Context-aware Migratory Services � Migratory Services Framework � Evaluation � Conclusions & Future Works 4

  5. Migratory Services Model Migratory Services Model MS migration Physical client-service interaction Virtual client-service interaction MS State n 5 MS State n 4 MS State Migratory C n 3 service Client end-point n 1 end-point Migratory Services Model - Migratory Services Model - cont cont’ ’d d � One-to-one mapping between MS clients and State n 5 MS migratory services State n 4 MS State Migratory C n 3 service Client end-point n 1 end-point MS S Meta-service State Migratory service n 2 5

  6. Example: Region monitoring Service Example: Region monitoring Service Migration MS MS Exit n 5 n 5 n 4 Migration MS n 4 MS n 3 Migration n 3 Response Response C MS S n 1 n 2 Request C n 1 Key ideas in Migratory Services Key ideas in Migratory Services � Capable of migrating to different nodes in the network in order to effectively accomplish their function � 3 basic mechanisms: • Monitor the context of interacting entities • Specify in context rules how the service execution is context-dependent • Migrate the service from node to node and resume its execution once migrated � Service migration is triggered by context changes � Service migration is transparent to the client 6

  7. Outline Outline � Motivations � Context-aware Migratory Services � Migratory Services Framework � Evaluation � Conclusions & Future Works Migratory Services Framework Migratory Services Framework 7

  8. Migratory Service Implementation Migratory Service Implementation using Smart Messages using Smart Messages � Smart Messages • implemented on a modified version of Sun’s Java K Virtual Machine � Migratory Services • clients, migratory services, and meta-services are Java programs that register with the framework • the framework maps these programs onto lower- level SMs • SM self-routes using geographical and content- based routing Context Manager Context Manager � Context data provided by the SM platform • location, time, speed using GPS • device status information • neighbors list � MonitoredCxt identifiers are translated into SM I/O tags � Access to context data by polling or blocking on corresponding SM tags 8

  9. Context Rules and Validator Validator Context Rules and � Evaluate if a service computation can be “correctly” carried out on the current hosting node � If not, trigger migration � CxtRules are service/client-specific policies • inCxtRules – control on incoming data • outCxtRules – control on outgoing data � CtxRules are condition/action statements • Conditions are full binary trees of Boolean expressions (comparisonNodes + combinationNodes) • Ex: < OR , <batteryLevel, EQUAL, low> , <responseLocation, OUT_REGION, userRegion> > • Actions: migrate service , send update , accept/refuse response ,... Communication Manager Communication Manager � Tasks: • Discover meta-services • Route messages between communicating end-points • Carry out service migration � Use naming conventions defined by SM � Two basic SM routing algorithms: • geographical routing (similar to GPSR) • region-bound content-based routing (similar to AODV) 9

  10. Reliability Manager Reliability Manager Active Inactive Client Service Service � Fault-tolerance to one Update Response failure Response Update Response � Inactive version of the Response Update Response service on a secondary Response node � In case of failure of the Timeout Request primary version, the Delete secondary version can Response take over the service Response Response provisioning primary secondary node node Outline Outline � Motivations � Context-aware Migratory Services � Migratory Services Framework � Evaluation � Conclusions & Future Works 10

  11. Prototype Migratory Service: TJam TJam Prototype Migratory Service: Predict traffic jams in � real-time Traffic jams are � locally congested phases in which cars travel at slow or zero velocity avg - min num num P = maxP × max number number TJam utilizes two � - min num num types of information avg - max speed speed P = maxP × min speed speed that every car owns: - max speed speed α α P' = × P +(1- )× P number of one-hop tjam number speed • neighboring cars N jam P = P' × N tjam tjam speed of one-hop total • neighboring cars TJam: : Testbed Testbed Experiments Experiments TJam � Ad hoc mobile network of Ad hoc mobile network of � 11 HP iPAQs with 802.11 11 HP iPAQs with 802.11 cards and GPS cards and GPS � Use mobility traces Use mobility traces � � 1 1- -2 hops communication 2 hops communication � � 2/3 neighbors 2/3 neighbors � 11

  12. TJam constantly executes in the user constantly executes in the user- - TJam specified region specified region 3600 3200 2800 2400 Location (m) 2000 1600 1200 user location service location/correct answer 800 service location/wrong answer updates 400 user range 0 0 50 100 150 200 250 300 350 400 time (sec) TJam: Simulations TJam : Simulations � ns-2 simulator with the CMU-wireless extensions � microscopic traffic generator tool Micro-VTG � Goal: Investigate the scalability of migratory services in large scale networks � Study based on the comparison of • TJam-Smart : migratory service model implementation • TJam-Base : baseline centralized approach � Metrics • inter-response time • correct response generation time • packet utilization rate • response packets overhead 12

  13. Simulations: effects of number of clients Simulations: effects of number of clients � highway of length 25km with 3 lanes � vehicles avg speed is 30m/s with a gap of 150m � 800 vehicles (500 vehicles active and 50 service nodes) Correct response generation time (Smart) Correct response generation time (Base) Inter-response time (Smart) Inter-response time (Base) 0,07 3,5 Inter-response time (sec) 0,06 3 generation time (sec) Correct response 0,05 2,5 0,04 2 0,03 1,5 0,02 1 0,01 0,5 0 0 50 100 150 200 250 300 Clients number Simulations: effects of vehicles speed Simulations: effects of vehicles speed � 150 clients Correct response generation time (Smart) Correct response generation time (Base) Inter-response time (Smart) Inter-response time (Base) 0,06 3 Inter-response time (sec) generation time (sec) 0,05 2,5 Correct response 0,04 2 0,03 1,5 0,02 1 0,01 0,5 0 0 10 20 30 Vehicles speed (meter/sec) 13

  14. Outline Outline � Motivations � Context-aware Migratory Services � Migratory Services Framework � Evaluation � Conclusions & Future Works Conclusions Conclusions � Migratory Services enables a new class of services in ad hoc networks • services quickly adapt to changes in the physical environment, in the node capabilities, and network topology • service continuity to the client � Experimental results demonstrate the feasibility of our approach � Simulation results demonstrate the scalability and efficiency of migratory services compared to a traditional centralized approach 14

Recommend


More recommend