Enterprise Application Integration & SAN - Technology Information

Computer Technology Review, Jan, 2000 by Fred Moore

By now, most or us nave read about, been involved with, or had direct experience with SANs (Storage Area Networks) based on the amount of market visibility given to the topic. Can we say the same about EAI (Enterprise Application Integration)? Let's take a look at these two emerging areas in order to understand their increasing value to each other.

EAI Middleware Overview

What is EAI? EAI is packaged software that provides the foundation for integrating enterprise software applications and business processes that run on different computer systems. EAI middleware solutions significantly cut the time and cost of development, integration, and maintenance of applications that exist on legacy, open systems, and distributed platforms. In the case of EAI, middleware is the software glue that helps databases and applications that run on different computer systems work together. Application integration, thus, includes any effort to enable to or more application systems to communicate and share each other's most valuable asset--their data. Analyst projections call for the EAI middleware revenue market to reach $600 million in 1999 and exceed $5 billion by year-end 2003, a CAGR of over 75%.

The EAI middleware market is being driven primarily by the following items: 1) The re-centralization of computing resources, 2) The need to access data from applications running on OS/390, Unix, NT, and Linux platforms, 3) The tremendous effort and expense to accomplish application integration through labor-intensive point-to-point programming and data migration or conversion methods, and 4) E-business, Internet, and .com paranoia. Relatively speaking, many customer environments today looking at EAI have significant investments in batch-oriented systems and modifying these systems is difficult, given the well-publicized shortage of software developers. Costs of modifying existing software applications are estimated to range from $2.50 to $4.00 per line of code!

Several trends led to developing distributed and client/server computing systems of the late 1980s and early 1990s in the attempt to break down what was perceived to be large and more costly centralized computing environments. Applications were developed in a variety of areas within a business and were seldom coordinated. As a result, islands of information were created with little thought given to the longer-term requirement to communicate and share information between the islands. The islands were often built around different computing systems with Unix, NT, and 0S/390 (then MVS) at the core.

By the mid-1990s, we knew that distributed computing and its management was often significantly more expensive, created redundancy, and was more labor intensive than managing resources centrally. This reality has fueled the present trend to centralization of computing resources with distributed access. In addition, changing applications in these systems was very difficult and costly. It is now estimated that 70% of business critical data is stored in OS/390 (the mainframe) class enterprise servers. This situation is not likely to change in the near future, given the high availability characteristics of these systems (now at 99.999% for Sysplex). The mainframes performance, stability, and scalability make it the most secure and reliable method for storing both historical or legacy data and current data files.

Enterprise or mainframe systems are now growing in number for the first time in the decade. Nonetheless, much of the tremendous IT growth is based on open systems platforms, typically NT or Unix platforms, as it is easier to develop new applications on these systems. Open systems, with their inherent support for and the standardization of TCP/IP networking, low cost, and easy implementation, are excellent choices for implementing the distributed architecture of the Internet or company Intranets. Networking between the various departments and organizations will become much easier.

By the way, XML (Extensible Markup Language) is quickly looking like the final piece of this basic network infrastructure. But what happens when data movement and shared file access becomes more intensive between the various departments within a business or between businesses (B2B) and outgrow the I/O capabilities of LAN-based data transfer? Unfortunately, the goal of real-time universal data access has, until now, been unattainable. Different implementations achieve part of the goal, but each misses in at least one critical criterion. The access is either not instantaneous or the data is not in its native mode. The system cannot directly access data wherever it resides on the network and, as a result, it must move the data from the core database and often reformat the entire file. Additionally, these solutions are expensive, proprietary, time-consuming, and resource intensive to implement.

Does the growing EAI market clearly identify an application area that SAN architecture has targeted? If one server did not need to access the information on another server, EAI wouldn't exist. Our highly integrated world of computing suggests that connecting all types of nodes on a worldwide network is no longer an option. Projections are now indicating that the number of IP addresses will match or exceed the world's population by 2005. NAS can provide a good solution in low to moderate data intensive systems. The EAI model of connecting unlike servers and their respective users to enable data access and data sharing nicely overlays the SAN model for allowing heterogeneous servers to access a common pool of storage.