ebook img

Dynamic Load Balancing in Distributed Content-based Publish/Subscribe by Alex KY Cheung A ... PDF

147 Pages·2006·0.82 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Dynamic Load Balancing in Distributed Content-based Publish/Subscribe by Alex KY Cheung A ...

Dynamic Load Balancing in Distributed Content-based Publish/Subscribe by Alex K. Y. Cheung A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Graduate Department of Electrical and Computer Engineering University of Toronto Copyright (cid:176)c 2006 by Alex K. Y. Cheung ii Abstract Dynamic Load Balancing in Distributed Content-based Publish/Subscribe Alex K. Y. Cheung Master of Applied Science Graduate Department of Electrical and Computer Engineering University of Toronto 2006 Distributed content-based publish/subscribe systems to date suffer from performance degrada- tion and poor scalability under load conditions typical in real-world applications. The reason for this shortcoming is due to the lack of a load balancing solution, which have rarely been studied in the context of publish/subscribe. This thesis proposes a load balancing solution specific for distributed content-based publish/subscribe systems that is distributed, dynamic, adaptive, transparent, andaccommodatesheterogeneity. Thesolutionconsistsofthreekeycon- tributions: a load balancing framework, a novel load estimation algorithm, and three offload strategies. Experimentalresultsshowthattheproposedloadbalancingsolutionisefficientwith less than 0.7% overhead, effective with at least 90% load estimation accuracy, and capable of loadbalancingwith100% ofload initiatedat anedge nodeof theentiresystemusing real-world data sets. iii Acknowledgements Super thanks to Mom, Dad, sister (Annie), friends (Mr. Biggy, Mr. Dum Dum, Mr. Chestnut) for their indirect support in my Master stud- ies. Thanks to KUG friends (Chef, Knifey, Chopsticky), and people in and outside the PADRES team (Eli, Guoli, Pengcheng, Serge, Vinod, and others) for their friendly companionship and guidance. Thanks to Cybermation for their funding in the PADRES project and their free all-you-can-eat snacks and drinks. Last but not least, big thank you to my supervisor, Professor Hans-Arno Jacobsen, for his unmatched super- vision and giving me a lot of freedom. Post defense: Thank you to my J.A.L. defense committee for giving an almighty A+ for this thesis. Yay! :D iv Contents List of Tables ix List of Figures xi List of Terminology xv 1 Introduction 1 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Background and Related Work 7 2.1 Publish/Subscribe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1.1 Publish/Subscribe Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1.2 Content-based Publish/Subscribe Language . . . . . . . . . . . . . . . . . 7 2.1.3 Content-based Publish/Subscribe Routing . . . . . . . . . . . . . . . . . . 10 2.1.4 Load in Publish/Subscribe . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.2 Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.2.1 Load Balancing in Different Software Layers . . . . . . . . . . . . . . . . . 12 2.2.2 Classification of Load Distribution Algorithms . . . . . . . . . . . . . . . 14 2.3 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 v 3 Load Balancing Framework 19 3.1 Overview of Load Balancing Components . . . . . . . . . . . . . . . . . . . . . . 19 3.2 Underlying Publish/Subscribe Architecture . . . . . . . . . . . . . . . . . . . . . 20 3.3 Load Detection Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.3.1 Protocol for Exchanging Load Information. . . . . . . . . . . . . . . . . . 23 3.3.2 Triggering Load Balancing Based on Detection Results . . . . . . . . . . . 28 3.4 Load Balancing Mediation Protocols with Brokers and Subscribers . . . . . . . . 35 3.4.1 Mediating Local Load Balancing . . . . . . . . . . . . . . . . . . . . . . . 37 3.4.2 Mediating Subscriber Migration. . . . . . . . . . . . . . . . . . . . . . . . 38 3.4.3 Mediating Global Load Balancing . . . . . . . . . . . . . . . . . . . . . . 39 4 Load Balancing Algorithm 41 4.1 Load Estimation Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.1.1 Estimating the Input and Output Load Requirements of Subscriptions . . 41 4.1.2 Matching Delay Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.1.3 Input Utilization Ratio Estimation . . . . . . . . . . . . . . . . . . . . . . 44 4.1.4 Output Utilization Ratio Estimation . . . . . . . . . . . . . . . . . . . . . 45 4.1.5 Memory Utilization Ratio Estimation . . . . . . . . . . . . . . . . . . . . 45 4.2 Offload Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.2.1 Input Offload Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.2.2 Match Offload Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.3 Output Offload Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 5 Experiments 69 5.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 5.2 Macro Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5.2.1 Local Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5.2.2 Global Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 5.3 Micro Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.3.1 Local Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 vi 5.3.2 Global Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 6 Conclusion 123 6.1 Summary and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Bibliography 127 vii viii List of Tables 4.1 Bit vector example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.2 Initial load of brokers B1 and B2. . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.3 Load of B1 and B2 after input load balancing. . . . . . . . . . . . . . . . . . . . 62 4.4 Load of B1 and B2 after a favorable case of output load balancing. . . . . . . . . 62 4.5 Load of B1 and B2 after an unfavorable case of output load balancing. . . . . . 63 5.1 Subscription templates used in all experiments. . . . . . . . . . . . . . . . . . . . 71 5.2 Default values of load balancing parameters used in experiments. . . . . . . . . . 73 5.3 Broker specifications in local load balancing experiment. . . . . . . . . . . . . . . 74 5.4 Broker specifications in global load balancing experiment. . . . . . . . . . . . . . 91 5.5 Edge-broker specifications in global load balancing micro experiment. . . . . . . 115 ix x

Description:
ent CPU speed, network bandwidth, memory size, and operating system. Subscriptions Its entire operation requires zero-human involvement. To date, very limited Meghdoot [15] presented the first evaluated load sharing scheme for Examples of previous work on topic-based publish/subscribe.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.