Penny-pinching IT managers should think twice about replacing a mainframe with servers, according to a recent study.
According to English research firm Xephon, Unix minis and PC servers cost about 2.2 times as much to operate
per user as their bulky brethren. Report author Barry Graham says when one factors in the various cost elements over five years, the difference becomes apparent.
The report divides costs into four categories: basic hardware and software and maintenance, application software, personnel costs and support. According to the data, the race was relatively close after the first categories were totaled (mainframes — US$5,500, Unix minis — US$6,500, PC servers — US$8,250). The gap further widens when the human element is added.
The report also makes reference to hidden costs. Servers provide response times in the two-to-four-second range, for example, which is two to three seconds slower than a mainframe. If a response is needed every 45 seconds, this equals a five per cent overhead. “”It is unlikely that end-users will be able to do any useful work during that time,”” the report states. “”At US$36,000 per person per year, this adds a minimum figure of US$1,800 a year to the Unix system costs, or US$9,000 over the five years.””
In terms of support staff costs, Xephon data suggests mainframes come out on top. According to the report a mainframe will cost US$1,500 per user for five years, US$4,500 for Unix minis, and US$9,000 for PC servers.
“”You sit there and think, ‘I can buy this NT server for $50,000. Surely it must be cheaper than the mainframe.’ What you don’t realize is in no time at all you’ve got 50 of these damn things and 25 guys trying to make them all work together,”” Graham says.
Framingham, Mass.-based research firm IDC completed a similar study in the mid-1990s and drew the same conclusion, according to Steve Josselyn. IDC’s research director, enterprise server fundamentals, Josselyn says it’s no secret Unix environments need more administrators than mainframes do.
“”Everybody was basically saying the mainframe was this huge sinkhole in terms of money and if you really compared it over a period of time, then it wasn’t as bad as people made it out to be–especially the Unix guys that were trying to penetrate that market,”” Josselyn says.
While costs metrics are often a hot topic of debate, Graham says he has total confidence in the numbers. “”We work with big clients negotiating their hardware and software contracts. So really we started from a point which is different to anybody else. We started from the point of actually knowing what these things really do cost.””
While Graham is adamant the totals are right, he says he isn’t trying to convince anyone mainframes are right for them.
“”We’re not trying to tell you which is best for your business. Nobody knows whether it’s best to have a distributed or centralized system for a given customers unless you talk to that customer,”” Graham says.
Gary Pinkerton, senior account manager at IT outsourcing firm Zycom Technology Inc.‘s Ottawa office, says buyers should remember this when looking for a system. He says there are a number of questions that need to be answered before making the leap because “”price is only one small factor.”” Customers need to know if they need the power of a mainframe, for example. Is central management a must-have? Are applications available?
“”It all depends on where the customer’s priority is and what business problems they’re trying to solve,”” he says. “”If they’re trying to solve a database management system then that’s a totally different kettle of fish (compared to desktops).””
Xephon first conducted research on mainframe costs in 1992 and has periodically updated its data. This is its first report in the subject in about five years.