There's a poster that anyone who has ever visited a computer operations center has probably seen. It shows a stylized dinosaur batting around, throttling, and devouring a bunch of smaller creatures--which, on close inspection, turn out to be personal computers. The slogan below is: "Mainframes: We're Back, and We're Pissed."
When I first saw this poster, I enjoyed the humor. As it persisted, I wondered at the wisdom of choosing a dinosaur--a universal symbol for extinction, as well as antiquity--as the alter ego for a retro-technology crusade.
The emotions behind the slogan come primarily from fear of change, and the very weight of Big Iron's history. Changing technology has put the PC/server and the mainframe/green-screen terminal on markedly divergent paths. People with 20 years invested in a particular computing environment would clearly rather call out names than switch--but are willing to switch when the mainframe jobs start to disappear.
Information-systems departments in some companies are loaded with staff whose glory days hark back to the IBM 370 and its contemporaries. Many managers with mainframe backgrounds have been forced to oversee the technicians and engineers who keep the machines running. Last week, a student told me that his employer's lead security engineer had stated that their proxy server was "a multiplexor."
This bias is particularly damaging when it spills over to support decisions. Two months ago, I was awakened to address a system failure; it turned out that the power supply of a server providing primary DNS name-resolution services had failed, and, due to a firewall that wasn't configured to access the still-functioning secondary DNS server, was affecting about 3,000 users.
The shame, however, was that the monitoring staffer at our data center had noticed the failure three hours earlier, but hadn't called anyone until after 5 a.m. because he was working on another failed application. He didn't call me at 2 a.m. because the other application--which affected, at most, 50 people--ran on the mainframe, and his manager had told him that mainframe applications always have priority. In fact, the DNS server's failure turned out to be the cause of that application's problem.
Mainframe hangers-on are not found in every company, however. They are rare in those that produce computer products. Big-iron advocates ascribe this to the fact that the majority of software and hardware makers focus on microcomputer platforms.
But the real gap occurs in the hyper-dynamic Web-sales sector of the economy, where mainframes are virtually unheard of. The bulk of the mainframe renaissance of the past few years has taken place in large enterprises that have been around for 30 years or more--in other words, with companies that already owned mainframes. The mainframes apparently referred to by the dino poster are back because some organizations have included legacy systems in the design of client-server applications.
For the most part, these applications' core processes execute on servers, but might pull data from a mainframe host's store. In essence, the mainframes have returned in the form of disk farms, and only at companies that had them around anyway and would otherwise probably discard them to save on electricity. Paradoxically, then, aren't mainframes back only because the server application developers found something to do with them?
It would be nice to end this with a plea for peace. But it would not be genuine, for the slowness of the shift toward PCs is not evidence that it isn't proceeding. Dinosaurs didn't die out overnight, but they are long gone nonetheless.
Contributing Editor Joe Rudich is a network administrator with the St. Paul Companies in St. Paul, Minn.
