Publisher Mobility in Distributed Publish Subscribe Systems Vinod Muthusamy

Publisher Mobility in Distributed Publish/Subscribe Systems Vinod Muthusamy, Milenko Petrovic, Dapeng Gao, Hans-Arno Jacobsen University of Toronto June 10, 2005 4 th International Workshop on Distributed Event-Based Systems (DEBS'05)

Motivation n Explosion of information producers q Blogs, wikis, podcasting, photo sharing n Mobility of users q Cell phones, PDAs, sensors n Mobile information producers q Traditionally wired publishers can increasingly be mobile q New types of publishers n n Pub/sub data dissemination q Well suited to mobile clients n n SMS, camera phones, location based services Decoupling, filtering Mobility of information producer has not been studied in pub/sub q Breaks common pub/sub assumption June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 2

Publisher Mobility Scenarios n Journalists with blogs q q n Update blogs on location Upload pictures from camera phone Police patrol car q Send status updates n n 2 Traffic, accidents, parts failures Publisher Mail delivery q 1 Track delivery status, location updates June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 3

Agenda n Background q q n Publisher mobility q q n Problem Solutions Evaluation q q n Context Subscriber mobility Setup Results Conclusions June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 4

Context n Part of Toronto Publish/Subscribe System (To. PSS) q Improve expressiveness n q Distributed issues n q Fault tolerance, load balance, reliability New environments n n Approximate matching, location queries, XML, RDF, composite subscriptions, historic subscriptions, etc. MANETs, P 2 P overlays, sensor networks Mobile-To. PSS project q Subscriber mobility [MDM’ 04] n q q q Based on JEDI, SIENA work Publisher mobility [DEBS’ 05] Effects of routing computations [Mobicom’ 05] Content based routing in MANET [Mobiquitous’ 05] June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 5

Distributed Publish/Subscribe n Advertisements flooded q n . . . Create adv tree Subscriptions along reverse adv path q n . . . Create multicast tree Publications along reverse sub path Subscriber Publisher Advertisements Subscriptions Publications June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 6

Subscriber Mobility Problem n n n Matching publications during disconnection q Stored by broker q Replayed upon reconnection “State” transfer is expensive q Double message load with only 10% of mobile subscribers [MDM’ 04] No state lost when publishers are disconnected q No problem with mobile publishers? June 10, 2005 (DEBS ’ 05) 1 Subscriber Mobile-To. PSS (University of Toronto) 2 Subscriber 7

Publisher Mobility Problem n n n q n . . . Adv and sub trees Moveout: both trees torn down Movein: both trees rebuilt Expensive q Network load: . . . 1 2 May be # ads > # subs No delivery until tree constructed Distinguish temporary disconnections t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker moveout t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) Publisher 8

Publisher Mobility Problem n n n q n . . . Adv and sub trees Moveout: both trees torn down Movein: both trees rebuilt Expensive q Network load: . . . 1 2 May be # ads > # subs No delivery until tree constructed Distinguish temporary disconnections t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker movein t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) Publisher 9

Prefetching Optimization n Exploits knowledge of future mobility patterns Concurrent q Construction at new broker q Teardown at old broker n . . . 1 2 Tree construction time hidden from user n t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker moveout t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) Publisher 10

Prefetching Optimization n Exploits knowledge of future mobility patterns Concurrent q Construction at new broker q Teardown at old broker n . . . 1 2 Tree construction time hidden from user n t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker movein t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) Publisher 11

Proxy Optimization n n Maintain trees from several brokers Advantageous if restricted mobility region t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker . . . 1 2 moveout movein Publisher t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) 12

Delayed Optimization n Maintain trees at old broker for some time Allow new tree to graft onto old tree Remove extraneous portions of old tree t 1 t 3 At Old Broker t 4 Disconnected t 5 At New Broker . . . 1 2 moveout movein Publisher t 2 Disconnect (moveout) June 10, 2005 (DEBS ’ 05) Connect Can publish (movein) new events Mobile-To. PSS (University of Toronto) 13

Evaluation: Setup n Simulation Environment q q n ns-2 network simulator Implemented mobility optimizations Parameters q Topology n n q q Subscribers: 500 Publishers: 50 Locality: random, 30%, 60%, 90% Mobility n n n Metropolitan Area Network 4 levels of degree 4 64 leaf brokers • • • 1 64 Static subscribers, mobile publishers Random speeds (5 km/h, 50 km/h, 100 km/h) Metrics q q Tree rebuild load Tree rebuild time, delivery ratio June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 14

Publisher Scalability n Standard and Prefetching >> Proxy and Delayed n Prefetching worse due to extra control messages n Delayed better due to smaller tree deltas June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 15

Publisher Scalability n Probe tree completion n Prefetching is fastest q Starts early Standard is slowest q Almost 4 s Delayed close to Prefetching n n n Note: time is not known to publisher June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 16

Publisher Scalability n Tree rebuilding cost q q n Tree rebuilding time q q n Best: Prefetching, Delayed Worst: Standard Prefetching q q n Best: Delayed, Proxy Worst: Standard, Prefetching Good for the user Bad for the network Delayed q q Good for user and network Practical June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 17

Publication Locality n 250 publishers n Vary publication similarity n Standard and Prefetching approach Proxy and Delayed June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 18

Publication Locality n Time from publish to notification n Again, Standard and Prefetching approach Proxy and Delayed June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 19

Publication Locality n With sufficient publication similarity, optimizations have diminishing benefit q q Tree rebuilding cost Delivery latency June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 20

Conclusions n The publish/subscribe model is well suited to mobile applications q n But publisher mobility has not been evaluated Publisher mobility is expensive q q Breaks conventional assumptions Tree rebuilding imposes large cost n q n Must distinguish temporary vs. permanent disconnection Delayed has best performance and is most practical Future Work q q Other scenarios: realistic traces, mobile subscribers Develop more optimizations June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 21

Publisher Mobility in Distributed Publish/Subscribe Systems Thank you June 10, 2005 (DEBS ’ 05) Mobile-To. PSS (University of Toronto) 22