200 likes | 312 Vues
ZigZag : An content-based publish/subscribe architecture for Human Networks. Yaxiong Zhao and Jie Wu Computer and Information Sciences Department Temple University. Outline. Introduction System design Content representation model Publish/subscribe routing Performance evaluation
E N D
ZigZag: An content-based publish/subscribe architecture for Human Networks Yaxiong Zhao and Jie Wu Computer and Information Sciences Department Temple University
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Human networks • A wireless network composed of hand-held devices • The hardware is becoming ubiquitous • Decent wireless communication capability • 3G/4G/WIFI/Zigbee/Bluetooth/Infrared etc. • Yet there is no software/protocol for efficient communication between devices • Without the aid of the infrastructure
Content-based publish/subscribe • Friendly to users • Interest-based communication without knowing the end-points • Privacy preserving • Simple to establish communication • For a wireless network • Don’t have an efficient protocol • This is the focus of this paper
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Boolean expression based content representation • A subscription is expressed as the conjunction of multiple primitive Boolean expressions • {brand = Nikon AND focal_length ≥ 50mm AND focal_length ≤ 105mm} • A subscription can be viewed as a hyper rectangle in a high-dimensional space • Dimension corresponding to the number of attributes
Efficient approximate content processing • Exact content processing consumes scarce resources • CPU power and memory • Affect the user’s experience • Energy • Reduce the life time of the hardware • Introducing approximate processing • Transform the subscription into a discrete structure • Good for efficient implementation • False positive rate (FPR) • The ratio of the falsely matched messages to the truly matched ones
Z-order curve-based transformation • A discretization approach using Z-order space filling curve • Simple to compute • The entire multi-dimensional space is transformed into discrete cells • Each subscription is represented all the overlapped small cells
Computing Z-order transformation • For Z-order curve, a cell’s ID is obtained by interleaving the bits of its coordinates’ binary forms • For example, the coordinates of cell A in Fig. 2 is [2; 2], or [10; 10] in binary form. • The corresponding ID after Z-order curve transformation is 12 or 1100 in binary form.
Cont’d • Cell IDs are compressed into intervals • Which represented by two end-points • Intervals are inserted into interval-tree • Standard data structure for O(logN) query
Bloom-filter-based compression • Put the transformed cell IDs into a bloom filter • Reduced storage • Faster query • Introducing additional false positive • Acceptable in real application • Evaluated in simulations
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Routing based on aging • Similar to the Age-based routing FRESH • Henri et al MobiHoc’03 • Forward the message to the users that have a smaller encounter age to the matched subscribers • By comparing the residual TTL of the subscriptions • Subscription’s TTL gets updated
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Additional FPR of using Bloom filter. 4 hash functions are used. The bit-vector uses the same amount of memory as the interval tree.
Outline • Introduction • System design • Content representation model • Publish/subscribe routing • Performance evaluation • Conclusion
Architecture design for future wireless networks • Z-order curve based subscription representation and processing • Bloom filter based subscription compression • Age-based message routing and forwarding protocol • Our future work will be a prototype system working on smart phones.
Q&A Thanks for listening