October 2000

Return On Relationship With Legacy Systems

BY TONY RYBCZYNSKI

The impacts of (re-)engineering the enterprise around ROR extend right across the organization. While dot-com companies have focused their businesses on a single channel (the Web) and have built applications and operational systems around that business model, most enterprises have multiple service delivery channels, including brick and mortar channels (branches, stores, clinics, and sales offices) and electronic channels (the telephone, fax, e-mail, and chat, as well as the Web).

One key challenge for most established enterprises is interfacing their Web-based e-business systems into legacy mainframe-based applications. The reality is that in spite of the computing evolution to the client/server model, 70 percent of business data and applications continue to be based on mainframes, which are dominated by IBM and their System Network Architecture (SNA).

THE LEGACY OF IBM'S SNA
Up until the mid-70s, the enterprise had no choice but to build application-specific networks (for point-of-sale applications, financial transaction networks, data entry, and batch processing) using a variety of protocols and systems. This was a world of dumb terminals and large, expensive mainframes.

At this time, the Internet was still an academic network, and Ethernet still hadn't been invented. Mainframe vendors (for example, IBM, DEC, and Burroughs) responded to this situation by introducing proprietary networking architectures. Most had varied lives, evolving over the years. And, in many enterprises, these architectures still exist, having assumed the status of legacy systems.

IBM's SNA stands out because it represents the largest installed base. A traditional SNA network consisted of a number of terminals connected to a cluster controller, and a number of controllers multidropped on a low-speed leased line off of a front-end processor, itself "channel attached" to the mainframe computer via a high-speed interface. While IBM owns the SNA specification, many vendors have established themselves as players in the SNA space.

Very briefly, the SNA architecture consists of a set of networking, session, and presentation protocols; a set of physical and logical node types representing different levels of intelligence and control (all the way from dumb 3270 terminals to "center of the universe" mainframes); and, of course, a set of acronyms and terminology. Manageability, reliability, and predictability were three key attributes of IBM SNA networks, and these attributes allowed SNA networks to become the workhorses of mission-critical enterprise networking in finance, retail, and transportation industries (to name a few).

Interestingly, the strengths of SNA networks happened to be the weaknesses of IP networks, which have made the marriage of these environments a real challenge. To illustrate this point, when LANs took off in the form of Ethernet, IBM developed its token ring LAN technology with better manageability, reliability, and predictability. Only relatively recently has IBM embraced Ethernet, which can now deliver token ring performance at a considerably lower cost, through the use of switching and quality of service capabilities. In the WAN, SNA networks for the longest time were and continue to be run in parallel to multiprotocol routed networks using frame relay networking technologies.

Generally, the marriage of SNA networking and IP networking has evolved in three distinct stages:

• Stage 1. In the 90s, a technology called Data Link Switching was developed, allowing consolidation of SNA traffic with other data traffic over an IP network. The typical configuration consisted of PCs emulating 3270 terminals connected to token ring LANs with a router providing the Data Link Switching adaptation functionality. The SNA traffic, now riding on IP, would be carried over the IP WAN to a router at the data center. This latter router would convert the traffic back to native SNA to be delivered over a token ring LAN to the front-end processor and to the mainframe application.
• Stage 2. PCs and servers emulating 3270 terminals would directly encapsulate the traffic into IP for transport, most often over Ethernet and a wide area router network. At the host end, various approaches were used, including software in the front-end processor or channel-attached gateways, to unencapsulate the date and transform it into SNA.
• Stage 3. PCs running standard browser software communicate with applications over IP networks. Various Web-to-host networking approaches exist. This market is accelerating rapidly due to end user selection of browsers as the preferred interface, and due to the strategic intent of enterprises to integrate e-business applications with data and applications residing in IBM host environment. However, this approach is not a panacea -- certainly not in transaction- and data-entry-intensive environments, which are common, for example, in the financial industry. Specialized (that is, non-browser-based) man-machine interfaces will continue to be used, often concurrently with browser-based solutions.

The Web-to-host market comprises products that run on servers or on IBM hosts and act as instant Web-enablers in front of IBM hosts. They are predominantly software products and are typically priced on a per-seat basis to match the terminal-based pricing of the past. The vast majority Web-to-host deployments have, to date, occurred within intranets for which there was a ready-made market for vendors. Concerns about security are addressed by the same features used in traditional host access products.

WEB-ENABLING MAINFRAMES FOR GREATER ROR
Maximizing ROR will require enterprises to integrate how they serve their customers across multiple service delivery channels, including "bricks and clicks." Highly scalable Web-to-host networking is a key requirement given the growing importance of the Web and the need to access legacy mainframe-based applications.

IDC predicts that Web access to the corporate mainframe will expand like never before as companies exploit opportunities in Internet commerce. Fueled initially by intranet implementation of Web-to-host access, followed in later years by implementation across extranet and the Internet, the market is expected to grow from $112 million in 1998 to $1.25 billion in 2003, with a CAGR of 62 percent. The evolution of traditional SNA networks to a Web-to-host environment represents a critical component of e-business solutions and a critical link in the chain of business transformation around the ROR model.

Tony Rybczynski is director of strategic marketing and technologies for Nortel Networks' Enterprise Solutions unit. E-mail questions or comments to [email protected].

[ Return To The October 2000 Table Of Contents ]