Moving Toward the Zero Latency Enterprise
Moving Toward the Zero Latency Enterprise
By: Rickland Hollar
Aug. 19, 2003 12:00 AM
The Internet makes it possible to deliver information almost instantaneously - anytime, anywhere - and is redefining the traditional boundaries around organizations and their IT systems. The Internet has turned buyers into sellers, sellers into buyers, and set new expectations for how services should be delivered. These expectations raise the bar for applications in terms of their need for interconnectivity and responsiveness. For businesses to remain competitive in this environment - or in the case of government agencies, responsive - they must embrace the idea that speed not only matters, but that it is now a key discriminator. Enterprises able to leverage the Internet's real-time nature and its technologies create competitive advantages that let them reap the benefits of greater efficiency, responsiveness, market share, and profitability. This promise has led businesses to look to interconnect their enterprise resource management (ERM), supply chain management (SCM), and customer relationship management (CRM) systems, both internally and externally; and government agencies to look for better ways to connect their systems with the public, their suppliers, and each other.
The Gartner Group coined the term "zero latency enterprise (ZLE)" to describe organizations that can exchange information with employees, trading partners and customers in near real time). The original focus was on internal systems, but much of what Gartner said applies equally to eliminating latency between internal and external systems. Figure 1 illustrates many of the concepts underpinning ZLE. In a ZLE organization, business events trigger system events that post actions and send responses throughout the enterprise. Bill Gates calls this level of interconnectivity a "digital nervous system." Like the human nervous system, the applications in a ZLE organization interconnect in such a way that they eliminate latency, which is the time gap between when the system receives information at one point and uses it, wherever needed, at others.
Becoming a ZLE
The first step in becoming a ZLE organization is identifying the major business processes within your organization that the ZLE architecture must support. One goal at this stage is to establish the architectural boundaries of the effort; will it deal with internal systems, external systems, or both? Another goal is to understand the dynamics of each business process: its tempo, meter, natural pauses and breaks. Understanding these dynamics is critical to recognizing latency and bounding the parameters for fixing it. It is important to remember that what constitutes latency in one process may be completely acceptable in another, even for the same application.
The next step is to decompose each business process into its applications and identify any latency points that exist. You should ask: What applications make up this process? Is latency a problem in either the overall process or one or more of its supporting applications? If so, how much of a problem? How much does its timing need to change? The result of asking and answering these questions is a list of business processes and applications that have latency problems.
Next you need to learn as much as possible about each latency point so that you can later devise techniques for removing, or at least minimizing, the latency. For internal systems, latency stems from several root causes (see Figure 2). Legacy systems are often stovepiped applications that were developed independently, over time, using different technologies. These applications create islands of information and functionality that are by their very nature difficult to integrate and share.
The same data belonging to different applications may be in different formats, follow different data validation and business rules, or be updated through completely different business processes. Interfaces within these older applications tend to be synchronous, tightly coupled, and driven more by the underlying technologies than the business needs they serve. Proprietary drivers, proprietary APIs, and proprietary formats represent only the tip of the iceberg when it comes to tying these systems together. One question you should ask is: What are each application's processing characteristics: batch, on-demand, or continuously running? Some may be batch oriented where you need them to be real time, others may have availability and reliability problems in cases where you need them to be 24x7. These issues frequently reflect age and technology differences that increase the difficulties in creating a coherent architecture.
Latency's causes multiply when you look at connecting internal and external systems. Each external system potentially represents a different set of technology, security, reliability, and manageability characteristics that your architecture must address.
Business process management is, in my opinion, the most important part of the architecture; it's the glue that ties applications together. It should reflect the enterprise's business processes: assembling, sequencing, and orchestrating applications to align them with the business's natural processes and work flows. A workflow manager, a rules engine, and collaborative tools can be critical components at this level. Employees, business partners, and customers should find easy-to-use, intuitive interfaces supporting your core business processes.
Data communications and routing in conjunction with business process management create the central nervous system for the ZLE architecture. Two fundamental architectures, shown in Figure 3, have evolved in this area: hub-and-spoke; and data, or information, bus. The hub-and-spoke architecture uses a central integration engine and message queuing products, such as MQSeries and MSMQ, to integrate across applications. In this architecture, applications deal with one another through the central hub; this is responsible for extracting, transforming, and routing data and coordinating activities throughout the overall system.
The information bus architecture takes a decentralized approach. This architecture implements a common messaging framework, frequently using a publish and subscribe model, for intercommunication. Applications connect to this bus through application adapters and pass messages to one another by placing them onto the bus. The information bus may use either a messaging or workflow manager to assist in routing messages. You can use either architecture internally; the information bus is clearly superior when connecting between internal and external systems.
XML has become the lingua franca for solving the data transformation and formatting problem. It provides a flexible, extensible syntax for expressing both information and its structure in a meaningful format. Legacy applications can apply Extensible Stylesheet Language Transformations (XSLTs) to XML documents to convert information within those documents into whatever format they need. Data transfer and replication tools are also available for extracting, transforming, cleansing, and loading data for those wanting to make minimum modifications to existing applications.
Application integration can occur at many different levels (see Figure 4). A key question is whether there is overlap in the data the applications process or the business rules they enforce. User interface integration integrates applications at the presentation layer. This level of integration is valuable for connecting independent applications into common business processes. Data integration integrates applications at the database level by copying, transferring, or replicating information from one data source to another. This is a good strategy when transfers are timely and business rules are sufficiently compatible. Business logic integration integrates applications' middle tiers, allowing each application to retain its original business rules and logic. This level of integration works best for tying existing, interdependent applications together into more streamlined processes. Component integration integrates applications through their application programming interfaces (APIs), common components, or function calls. Integration at this level may require you to write proxy interfaces for some components; change call interfaces from direct to RPC for others; or adopt a distributed object model such as DCOM, CORBA, or Web services. This integration form is most useful for creating components several applications or processes can share.
A critical part of the application integration analysis is looking closely at each latency point to determine both the level of integration and corrections needed. The first step is to identify the appropriate integration level for each application: presentation, business logic, or data. Simply changing the application's invocation characteristics may be enough to also change its latency characteristics for some applications.
In situations where that is not the case, the next step is to drill down into the application and its interfaces with an eye towards improving the application's performance characteristics. The first, and simplest, corrective measure is to identify and remove any inefficiencies or chokepoints within the application. A second option is to look at overlapping the application's processing with that of others by making it an asynchronous process. Making an application asynchronous is straight-forward; you simply need to add a queue and alerting and rendezvous mechanisms. This can also be a good approach for dealing with reliability and availability problems caused by older systems. A third, and sometimes only, option is to redesign and rewrite the application.
You may need to make several passes through each of the four architectural elements to finalize the ZLE architecture. That isn't unusual. It's important that you come away with an overall architectural strategy, a list of integration points, and an idea of the integration strategies you'll need to address as part of the implementation process, which is the next step. Before proceeding to the implementation phase, it's a good idea to create a set of guiding principles to help in making architectural tradeoffs and selecting products. Questions you should answer include: Are the number or types of products you use of concern? How about the amount of code you write? Is it important to use the same solution for solving the latency problem between both internal and external applications? Is it important to use the same integration solution for integration points at the same level? Do you have large investments in ERM, SCM, or CRM solutions that will drive the implementation? With the answers to these questions in hand, you're now ready to look at options for implementing the architecture.
XML, which is the heart of Web services, provides a standard for data representation. XSLT adds a language for data transformation and formatting. SOAP-RPC contributes a lightweight, standards-based, platform-independent component model for implementing distributed components. In short, Web services provides all the elements necessary to implement whatever ZLE architecture you ultimately develop. With several of the standards still evolving, the issue is that products lag behind standards; that means you have to write more code.
If that is a concern, off-the-shelf enterprise application integration (EAI) products offer a good foundation for moving towards a ZLE organization. EAI products provide message broker and adapter technologies that quickly integrate applications to exchange and share information at the data, business logic, or presentation layers. Most EAI solutions implement either a message broker or bus concept corresponding to the hub-and-spoke and information bus architectures. If you decide this is the best approach for you, choose a product that fits into your overall integration strategy by providing the greatest number of integration adapters corresponding to the integration levels, points, and products you identified as part of your analysis.
EAI and Web services are extremely powerful together - EAI for fine-grained interfaces, Web services for coarse-grained interfaces. Many EAI vendors, such as SeeBeyond, TIBCO, webMethods, and IBM, recognize this synergy and offer products that are in fact a marriage between traditional EAI technologies and Web services. These products give you a best of both worlds option. Ultimately, the question boils down to which strategy works best with your architecture within your organization.
Reader Feedback: Page 1 of 1
SOA World Latest Stories
Subscribe to the World's Most Powerful Newsletters
Subscribe to Our Rss Feeds & Get Your SYS-CON News Live!
SYS-CON Featured Whitepapers
Most Read This Week