Issue No. 02 - February (2007 vol. 8)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MDSO.2007.8
Gero M? , Berlin University of Technology
Ludger Fiege , Siemens Corporate Technology
Peter Pietzuch , Imperial College London
Over the past couple of years, event-based systems have appeared in many application domains, such as enterprise management systems, large-scale data dissemination, Internet applications, and autonomic computing. Event-based techniques have established themselves as an efficient way to structure such systems and handle challenging interaction patterns between components. However, event-based-systems research has been scattered across several conferences and workshops, making it hard to establish a single community of interested researchers and practitioners.
To support event-based systems' growth, we're organizing the Inaugural International Conference on Distributed Event-Based Systems in Toronto in June this year. The DEBS conference builds on the success of the five DEBS workshops held from 2002 to 2006 in conjunction with major conferences such as the International Conference on Distributed Computing Systems, the International Conference on Software Engineering, and the ACM SIGMOD International Conference on Management Data. By establishing DEBS as an independent conference, we hope to reach a broader audience from academia, industry, and standardization bodies. Here we look at why we believe the time is ripe for a DEBS conference.
Events in industry
Events and messaging have been in use in industry for years. With the advent of service-oriented and event-driven architectures, however, the eventing paradigm becomes a central building block of business IT.
Traditionally, events have been used, for example, in enterprise application integration to mediate business events such as orders and invoices that are documents and artifacts of business processes. At the same time, dedicated information-dissemination applications, such as stock quotes and network monitoring, have operated at a distinct, lower level of granularity. However, for some time, we've been witnessing a shift from specific applications to eventing as a general architecture paradigm. The zero-latency enterprise was propagated a while ago only to be superseded more recently by service-oriented and event-driven architectures (EDAs).
The objectives behind these changes are still the same: improving time-to-market and handling the competitive pressure to adapt to customer and market needs. This requires flexibility that traditional architectures don't really provide. Classic enterprise application integration focused on consolidating documents and data, but it has turned out that without changing the applications themselves, we can't make the application landscape more agile. The building blocks must be designed to provide value to others—they must be service oriented. Prime examples are companies such as Google and Amazon that show us this approach's possible impact and the need to be ready to integrate and offer new services to customers.
Service-oriented architectures don't prosper just because we communicate over HTTP. The components must be designed to serve by making them autonomous, composable, and responsive to input, exceptions, and changes. Event-driven design offers the flexibility to address these issues and reflects the real world's event-driven nature. The Gartner consultancy, among others, expects increasing interest in EDA. According to their hype cycle prediction, EDA is in a discussion phase before its broad adoption in industry will begin. 1 With the DEBS conference, we aim to provide a forum for this discussion on the principles, tools, and standards needed to implement EDAs.
Events in research
Event-based systems have proven a fruitful research area. When addressing some of the issues that industry is facing, computer science researchers have encountered fundamental challenges spanning the areas of distributed systems, databases, security, performance modeling, algorithm design, and programming models. Research on these topics could lead to particularly satisfying outcomes that directly affect industry. Academic research in event-based systems should exploit this golden opportunity, and we hope the DEBS conference will encourage this.
One area that would benefit from academic input is the provisioning of distributed publish/subscribe systems. Recently, we've witnessed increasing interest in systems that perform distributed content-based event filtering to support Internet-scale applications such as global RSS feed aggregation and filtering. Although such systems have led to successful research prototypes, so far they've gained little traction in the commercial sector.
One explanation for this might be the lack of techniques for provisioning the deployment of distributed pub/sub systems. It's difficult for any company to rely on a system architecture that isn't fully understood in terms of its scalability and performance characteristics when deployed across multiple data centers. As the number of users increases, the deployment must also scale up to cope with the higher demand. Up to this point, the research community has largely ignored these questions. This presents an opportunity for researchers to devise new performance models and provisioning techniques for content-based pub/sub systems. Such models and techniques could directly result in more commercial pub/sub deployments and enable the design and deployment of global-scale event-based applications that were inconceivable only a few years ago.
Events in standards
Several industry standards related to event-based programming exist, such as the CORBA Notification Service, 2 the Data Distribution Service, 3 and the Java Message Service. 4 Although these standards have gained some attention and have been used in practice, they haven't led to widespread adoption of an event-based interaction style.
With the advent of Web services, standardization efforts have begun to establish Web services in conjunction with related standards (such as XML) as a meta-middleware connecting applications running on various platforms. In many aspects, these efforts resemble past efforts to standardize middleware platforms such as CORBA. One crucial difference is that now all major players, such as Microsoft, IBM, TIBCO, and Google, are cooperating. Recently, these standardization efforts have also targeted event-based communication, 5,6 leading to open specifications such as WS Eventing 7 and WS Notification http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=wsn. This development is also pushed by the hype surrounding EDA and helps bring event-based programming to a broader public. The resulting momentum might lead to a breakthrough for event-based programming at a broader scale because companies would have to move ahead to keep from falling behind their competitors. Thus, it's becoming more important than ever for practitioners and researchers to contribute their knowledge and experience to this process.
The community seems to have recognized this, judging by the recent calls for participation in the workgroups of the Event Processing Technical Society. This group aims to bring together participants from industry and research. This year's DEBS conference will try to encourage this development by having prominent keynote speakers from industry and research and by seeking research and industry papers. The DEBS demo track also provides the opportunity to present working research prototypes and innovative products stimulating the establishment of a united community.
After years of being caught in specialized niches, the event-based paradigm is now a hot topic in many areas, ranging from sensor networks over Web services to Internet-scale applications. It's therefore important for industry and research to join forces and work together on novel technologies and standards. We hope that the DEBS conference will foster this development and stimulate the free exchange of ideas. With your help, DEBS will serve as a focal point to this vital, fast-growing community and attract submissions and participation from a broad audience. More information about DEBS 07 is at http://debs.msrg.utoronto.ca. The submission deadline for research, industry, and demo papers is 14 March 2007.
Gero Muehl is a postdoctoral researcher at Berlin University of Technology. Contact him at firstname.lastname@example.org