The Perfect BalanceA message brokering approach may be the best way to give IT decision makers the realtime information they need to manage services efficiently
By Kemal Delic, Laurent Douillet, & Umeshwar Dayal Continued from Page 1 For example, consider a typical situation faced by the operations manager, in which the overall end-user response time deteriorates because of infrastructure problems (databases, servers, networks, applications); failure of e-services delivery components (networks, Web content servers, payment systems, shipping systems); and rapid usage growth. These causes could also be intermixed, masking the root cause of the problem or complicating proper diagnosis and decision-making. In this case, the Decision-Support Console could track realtime compliance with the service level agreement (SLA), calculate the best possible compliance that can be achieved in the time remaining, and project future compliance (see Figure 2). Detecting an SLA failure, even if the conformance period has not yet ended, would allow the operations manager to take action such as prioritizing resolution efforts, assigning new support resources, and avoiding incoming traffic and service deterioration. Message Broker ManagementDesigning the management platform for the outsourcing business is a hugely complex task involving several difficult requirements as well as intricate technical challenges. An outsourcing vendor usually serves multiple enterprise customers with varying and (sometimes) contradicting requirements. Although the successful vendor will clearly address the majority of the requirements for responsiveness, scalability, reliability, and security, it will be very difficult to satisfy all these requirements simultaneously. We will address here only responsiveness to illustrate the importance of realtime decision support in a realistic situation. The Decision-Support Console in Figure 2 illustrates a characteristic situation in which data and information from different operational systems are fused into an operational portal view. News and various internal and external business tickers are integrated into the view. Summaries from various communication channels (phone, voicemail, and pager) are shown as well. Instant communication enables quick inquiry and reaction to the ongoing operational subjects. This view is rendered through a message brokering architecture setup in which we suppose, for illustrative purposes, a business situation arises in which two key customer accounts are having problems: One account shows a forthcoming violation of an SLA limit, and another account is experiencing network congestion. (See Figure 3.) These two situations will create an inflow of events (1) from the infrastructure, which will be checked against stored business logic thresholds, and the necessary information about the customer account (2) retrieved. Directory information (3) will be used to combine stored knowledge (4) about customer procedures and problems into a coherent view (5) proposing standard remedies. On confirmation, a case will be opened in a workflow system (6), which will trigger several actions and activities (7). Instant messages will be sent to account managers, emails dispatched to an analyst, and calls will be placed to affected customers if necessary. Therefore, incoming problems will be detected in timely fashion, and a set of coordinated actions will be taken to prevent further deterioration of services. At the lowest level of the architecture, heterogeneous and distributed data, applications, and business processes must be integrated. Everyone intuitively understands the importance of integration, but although several enterprises have tested the concept, only a few have accomplished the goal successfully in production environments. For example, one major airline company is using this message brokering approach to handle as many as 12 million events per day, and a major shipping and delivery company has integrated its whole business using the same approach. There are two main architectural directions here: request brokering (such as CORBA) or message brokering. CORBA is based primarily on the synchronous request-response paradigm, which is well suited to the domain of transactions. The asynchronous message brokering approach deals better with the world of events. As realtime decision support is more about event management, this approach has some advantages, although it remains to be validated in more realtime situations. In this approach, a message broker (MB) serves as a mediator between various message (or event) publishers and subscribers, enabling the effective integration of heterogeneous components (whether they are databases, realtime data sources, resources, applications, or complete subsystems). It provides asynchronous queues for a variety of sources (network, middleware, and system) and resources (objects, database, and applications) and can be modeled as a service provider with two principal activities: publishing and subscribing. Because this approach does not require applications to be session-connected, it directly affects scalability of the message brokering architecture. Furthermore, it enables many-to-many and any-to-any relationships, which makes it easy for you to connect diverse applications by publishing and subscribing to relevant messages (or events). The MB provides "intelligent" routing, message transformation services, rule processing, message warehousing, directory services, adapters, and APIs. As such, it represents the central point of the integration paradigm. Intelligent routing means that the MB can understand who the originator of the message is, and it may transform the message on the fly according to predefined rules and forward the message based on the proper routing information. Message warehousing implies persistency features of the MB (message archiving, auditing, integrity checking, and mining, for example). Thus, if the MB goes down, the message warehouse will still be preserved. Directory services are a natural part of the MB as it must know how to locate various systems using its services. Adapters (generally) are layers between the MB interface and source and target applications, and they contain application-specific logic for translating data formats between the source and the target data models, and for invoking the target application through interfaces provided by the application. APIs are typically defined as the means of direct access to internal MB services. The brokering paradigm with an MB serving as the central integration axis is the best way to build realtime systems for the large-scale outsourcing domains. Without an MB system, it may take humans several hours to react. The message brokering-based architecture, in contrast, provides relevant data and information within a couple of minutes, thus enabling humans to shorten their decision-making cycle significantly. In addition to this time contraction, more data of better quality has been collected as well. Bearing the MessageThe principal impact from deploying MB systems is quicker turnaround time from the triggering events in the managed domain to the reaction of the system or personnel with the actions and remedies. Besides shortening the decision-making cycle, this architecture integrates the outsourcing domain and maps the low-level infrastructure events and service events into high-level business events and decisions. Considering that some of the hosted businesses are critical and that downtime is expensive or nearly prohibitive, the specific architecture of the management platform gives some distinct advantages inasmuch as human decision-making is supported by the efficient IT system. Two difficult problems may lead to interesting future research issues: how to improve the efficiency of the view rendering and how to devise better business metrics for the outsourcing domain. Our experience has shown that between seven and 10 operational parameters are enough to monitor the performance of one subsystem (a help desk, for example), and these, in turn, drive five to seven financial indicators, enabling precise trending of business parameters. However, which operational parameters and financial indicators to compute for the overall outsourcing business is not well understood. When these are defined, research will be needed to develop efficient incremental algorithms for computing and refreshing these metrics. Given the large volumes and inflow rates of data involved with widely distributed data sources, this problem is a difficult one. Kemal Delic [kemal_delic@hp.com] is a lab scientist with Hewlett-Packard's operations R&D and a senior enterprise architect with relevant experience in knowledge management, Bayesian nets modeling, and realtime intelligent systems. Laurent Douillet [laurent_douillet@hp.com] is lab technical lead within HP's operations R&D division. He has eight years of experience in designing software solutions for the knowledge management and outsourcing domains. Umeshwar Dayal [umeshwar_dayal@hp.com] is principal laboratory scientist with HP Laboratories, where he leads a team performing research in data mining, knowledge management, and business process management. RESOURCESRanadive, Vivek. The Power of Now: How Winning Companies Sense and Respond to Change Using Realtime Technology. McGraw-Hill, Sept. 1999 Wiederhold, Gio, and Michael Genesereth. "The Conceptual Basis for Mediation Services." IEEE Expert, Sept./Oct. 1997 Related Articles at IntelligentEnterprise.com: "The Medium Is the Message," May 7, 2001 "IT: know Thyself," May 15, 2001
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| ||||||||||||||||||||||||||||||||
