Publish/Subscribe Event Systems and DHTs
From P2P Wiki
Contents |
Introduction
Publish/Subscribe (pub/sub) systems are characterized by two main actors. Subscribers are those actors in the system that are interested in receiving significant information. They employ subscriptions to define all their interests, related to the content of such information. Additionally, publishers are those actors who produce information (a)periodically. Usually, the literature denotes such pieces of information as events. For instance, we can see sensor networks from industry as pub/sub systems. Sensors become publishers, sending periodically information about temperature, fluid preasure and other important measurements. Instead, central control systems become the subscribers receiving such interesting measurements.
As a main classifaction, we can classify pub/sub systems as topic-based or content-based. Topic-based pub/sub systems are those from where publishers produce events related to some subject, usually called topic, and subscribers subscribe to these topics. This category of pub/sub systems are also called channel-based and are quite easy to develop. From the other hand, content-based pub/sub systems work in a similar way than those topic-based, but subscriber can define same values of interest to match with new events. For instance, subscriptions in this way can especify exact values (e.g. X==9) or range of values (e.g. X<4 or 2<X<90). It is easy to see that this kind of systems is more powerful, but at the same time, it is more complex to develop.
Related Work
Topic-based pub/sub systems
Work in progress...
Content-based pub/sub systems
There are various ways to perform subscriptions in a pub/sub system. We classify them mainly into the following techniques: covering [11], epidemic-based [2, 15], summarization [5, 13], source-based [17] and rendezvousbased [1,3,7,14]. Most of them define specific pub/sub networks, in some cases on top of peer-to-peer (p2p) networks. In some cases, the pub/sub system is accompanied by advertisements from publishers in order to meet both subscriptions and events [5]. PastryStrings [1] is rendezvous-based system and for a M-dimensional event it needs to contact to M independent trees.
Some works [6, 8] employ the technique of rendezvous to meet both subscriptions and advertisements. They have in common that each kind of event has its own rendezvous through the overlay network. But, on the one hand, [6] works with single-attribute events at a time, embedding subscription information into the node Id keyspace, but specifically for dynamic filters. Hermes [8] has to maintain routing information for both subscriptions and advertisements.
Baldoni et al. in their work [3] propose different rendezvous-based subscription techniques over Chord [12]. Specifically, they propose attribute-split, keyspace-split and selective-attribute mappings to convert subscriptions and events into one or more keys. Besides, they implement a multicast overlay primitive m-cast() into Chord and they employ m-cast() to forward such keys as necessary.
References
[1] I. Aekaterinidis and P. Triantafillou. PastryStrings: A comprehensive content-based publish/subscribe dht network. link In Proc. ICDCS’06, pages 23–23, 2006.
[2] E. Anceaume, M. Gradinariu, A. K. Datta, G. Simon, and A. Virgillito. A semantic overlay for self-* peer-to-peer publish/ subscribe. link In Proc. ICDCS ’06, page 22, 2006.
[3] R. Baldoni, C. Marchetti, A. Virgillito, and R. Vitenberg. Content-based publish-subscribe over structured overlay networks. link In Proc. ICDCS’05, pages 437–446, 2005.
[4] J. L. Bentley. Multidimensional binary search trees used for associative searching. [] Commun. ACM, 18(9):509–517, 1975.
[5] A. Carzaniga, D. S. Rosenblum, and A. L.Wolf. Design and evaluation of a wide-area event notification service. link ACM Transactions on Computer Systems, 19(3):332–383, 2001.
[6] L. Chen and G. Agha. State aware data dissemination over structured overlays. link In Proc. IEEE P2P’06, pages 145–152, 2006.
[7] A. Gupta, O. D. Sahin, D. Agrawal, and A. E. Abbadi. Meghdoot: content-based publish/subscribe over p2p networks. link In Proc. ACMMiddleware’04, pages 254–273, 2004.
[8] P. R. Pietzuch and J. Bacon. Peer-to-peer overlay broker networks in an event-based middleware. link In Proc. DEBS’03, pages 1–8, 2003.
[9] J. Pujol Ahulló, P. García López, M. Sànchez Artigas, and A. F. Gómez Skarmeta. SQS: Similarity Query Scheme for Peer-to-Peer Databases. In Proc. ISCC’07, pages 100–105, July 2007.
[10] A. I. T. Rowstron and P. Druschel. Pastry: Scalable, decentralized object location, and routing for large-scale peer-to-peer systems. link In Proc. IFIP/ACM International Conference on Distributed Systems Platforms, volume 2218, pages 329– 350, November 2001.
[11] Z. Shen and S. Tirthapura. Approximate Covering Detection among Content-Based Subscriptions Using Space Filling Curves. link In Proc. ICDCS’07, page 2, June 2007.
[12] I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan. Chord: A scalable peer-to-peer lookup service for internet applications. link In Proc. SIGCOMM ’01, pages 149–160, 2001.
[13] P. Triantafillou and A. Economides. Subscription summarization: A new paradigm for efficient publish/subscribe systems. In Proc. ICDCS’04, 2004.
[14] P. Triantafillow and I. Aekaterinidis. Content-based publish subscribe over structured p2p networks. [1] In Proc. DEBS’04, pages 104–109, May 2004.
[15] S. Voulgaris, E. Rivi`ere, A.-M. Kermarrec, and M. van Steen. Sub-2-sub: Self-organizing content-based publish subscribe for dynamic large scale collaborative networks. link In IPTPS’06, February 2006.
[16] B. Y. Zhao, J. D. Kubiatowicz, and A. D. Joseph. Tapestry: An infrastructure for fault-tolerant wide-area location and routing. link Technical Report UCB/CSD-01-1141, University of California at Berkeley, Berkeley, CA, USA, April 2001.
[17] S. Q. Zhuang, B. Y. Zhao, A. D. Joseph, R. H. Katz, and J. D. Kubiatowicz. Bayeux: an architecture for scalable and fault-tolerant wide-area data dissemination. link In Proc. NOSSDAV’ 01, pages 11–20. ACM, June 2001.
