Obstacles along the way - corporate connectivity

Software Magazine, May, 1988 by Larry R. DeBoever

OBSTACLES ALONG THE WAY

Like kids in a candy story, MIS managers faced with a connectivity problem have a bewildering range of alternate solutions.

For example, there are over 70 LAN manufacturers in the U.S. alone (although we are in a period of consolidation with 3COM/Bridge, Tandom/UB, DCA/Fox, etc.), and almost 600 micro-mainframe linke products. The same is true for virtually any other product or technology area.

This range of options is an obstacle to corporate connectivity, which is being sought by major companies. Almost 60% of large companies have a connectivity strategy, but the very real political, management and technical obstacles must be recognized.

The prolific connectivity product development is being fueled by the following facts:

* the size of the market supports research in basic technology;

* the slow emergence of standards and early maturity of the market permit new technologies to receive immediate attention;

* the need to utilize existing products for new requirements has created major opportunities for niche products; and

* traditional data communications product vendors move to defend their market share by providing a rich stream of product enhancements.

Vendors in each area must be aggressive in developing product features or positioning statements. Many connectivity problems may best be solved with a combination of products. Additionally, there are now "software" solutions to some problems typically viewed as the domain of hardware suppliers, and vice versa.

PRESERVE EXISTING INVESTMENT

The investment of the typical medium-to-large organization in data processing resources in huge. The "visible" investment is evident--equipment, staff and facilities. The hidden investment in user knowledge, application costs and information documentation is much more difficult to identify and calculate.

The interest in integrating new technology with existing investment has spawned hundreds of new products--niche products. The best examples of this are gateways and PC coax boards (e.g. Irma).

MIS management realizes that standards are essential for computer connectivity--not just physical connectivity standards but also logical connectivity standards such as peer-to-peer communications. Consequently, MIS management is increasingly unwilling to add other data communications technology until it rationalizes its existing investment.

Standards are only "usable" to an MIS manager if they have been implemented by the computer manufacturers to which the MIS manager has a significant investment. Therefore, a de facto standard implemented by major vendors, and supported by most other vendors, is more important than a formal Ansi, IEEE or CCITT standard which is not implemented by vendors.

Multiple standards are acceptable and may even be desirable to the MIS manager when particular tasks are suited to particular technologies (e.g. MAP standard). This is especially true if there are adequate gateways between standards (e.g. IEEE 802.3 to IEEE 802.5).

Formal standards must be developed, promulgated and implemented by vendors before they can have a market impact. For connectivity to be achieved in a complex environment, most vendors must support the same standard (often the same version of that standard), to achieve interoperability.

It appears as if the MIS manager is caught in a "Catch-22." The development of a new standard often takes years. And vendors tend to implement the standard only after the standard specifications have stabilized and a market has emerged. Of course, the market cannot emerge without demonstrable products. And until support from vendors is assured, or a gateway is available, MIS will often not embrace a new standard. This, in turn, slows vendor implementations.

An example is the slow emergence of IBM's System Network Architecture (SNA), announced in 1973. Other vendor implementation efforts did not begin in earnest until 1980 (including needed efforts by IBM).

PERFORMANCE ISSUES

Despite the general maturity of the technology to physically interconnect disparate computers, there are several areas which require high-speed connectivity, but which either high-speed connectivity products are not available or are only available at unattractive prices.

File transfer is probably the most common example of where this dilemma occurs. IBM has made significant progress here: for example, the addition of T-1 streaming to the 3725; T-1 support on the channel; and SDLC's support for larger blocks of data. But, when one looks at the installed base, we find that a majority of companies "default" on these options and rely on 1,920 blocks.

For our example, assume we wish to transfer a 350,000-character spreadsheet file from an IBM PC to an IBM 3081 via a remote 3274 controller. The PC is required with a PC coax board of the type discussed earlier which allows the PC to appear to the 3274 to be a 3278-like terminal. The 3274 is connected to an IBM 3275 FEP via a leased line that runs at 9600 baud.

(I would estimate that 50% to 60% of the PCs connected to IBM mainframes are connected this way.)