As the U.S. electric power industry unbundles, the industry and its regulators grapple with two big questions concerning the degree to which distribution services should be unbundled. First, what...
Moving Off the Mainframe
computer software purchased from a commercial source or another utility can be tested, interfaced, and converted. The package can be installed "as is." You adopt certain functions and processes, and revise other procedures within your company to fit the new package. At the other end of the spectrum, you can modify the package to meet the specific needs of your company before testing and conversion. These options affect cost, depending on how well all the pieces fit. In practice though, it can prove difficult to gauge the "fit" before choosing the system. Being wrong can have major consequences.
Building a new custom-designed CIS to fit company needs is almost always the slowest and most costly alternative. Custom design calls for important decisions up front. Should we change all at once, piece by piece, or function by function? Which hardware platform? Do we run on a mainframe (already proven) or convert to "client/server" (unproven) technology?
If you want to move off the mainframe but retain as much as you can of your old system and codes, you can simply "re-platform" or "re-host" your existing CIS. This approach usually combines a single, heavy-duty computer with a UNIX operating system. Or, you can re-design your CIS from the ground up using parallel server processing technology.
Given past experience in the utility industry, a stand-alone CIS mainframe fully configured for a large utility will probably cost about $10 million. That figure might buy 200-250 MIPS (millions of instructions per second) and 1,000 gigabytes (GB) of disk storage (one gigabyte equals one billion characters). To buy 4,000 MIPS and 1,000 GB with fully configured software in a client/server environment might run as low as $1 million dollars (em a 10-to-1 advantage.
Software licenses in today's mainframe world are significant. For example, a utility can expect to pay somewhere around $3 million per year in system software license fees. Hardware maintenance may run about 8 percent of the purchase price annually. For the hypothetical mainframe proposed above, this would mean $800,000 a year over the life of the system. However, in a nonmainframe environment, a plan to gradually renew hardware by replacing the most outdated components with modern versions would seem prudent. For example, a utility might plan to replace the whole system every five years, maintaining the newest technologies, at a cost of about $200,000 each year.
But the greatest advantage in moving off the mainframe lies in human resources. Yes, workstations cost money. But whenever open, modular, downsized client/ server technologies are introduced, productivity climbs. Energy and enthusiasm start to rise. Today's information-processing professionals no longer want to be mired in mainframe-based systems. With new challenges, new things to learn, plus systems that look good and feel good to users, you have the main ingredients needed to build job satisfaction.
For many years and at many companies, CIS conversions have meant rapid "Big Bang" conversions. Taking a single long weekend to convert the entire CIS usually involves the least cost, the fewest risks. But what if there isn't a weekend long enough