The Trouble with Tribbles
How in the world, you're wondering, am I going to link Tribbles to mainframes? Well, in less skilled hands it would definitely be a challenge. Ok, think it through with me...what is the most notable attribute of a tribble? Aside from small and furry, they multiply...quickly. And what, in our little raised-floor world, multiplies quickly? Small, furry servers.
I'm thinking back over a few mainframe shops that I'm familiar with that have moved off of the mainframe. One, County A, took 18 years. Another, School District B, took 12 years. (School District B invited me to the IBM-manual burning party when it was over). Even a large shop spent 8 years moving one application off the mainframe, and the database is still living on the mainframe, and being replicated. And one even wrote off millions of dollars of accounts payable rather than convert the data.
I have to wonder: Did these shops ever really save any money? And now, with the benefit of hindsight, are people making better decisions? As poet and philosopher George Santayana said: "Those who cannot remember the past are condemned to repeat it." I think you not only have to remember history, you have to acknowledge that it happened. There are thousands of ways to rationalize anything you want to do in this day and age.
As I watched these 'modernizations' taking place, and saw them comparing apples to oranges, I was repeatedly stunned by the group-think, airline-magazine decisions. I'm tempted to go into Jo Galloway's tirade from A Few Good Men: 'in the same fast-food, slick-ass persian bazaar manner". It almost seems as though it was all one big conspiracy theory: "Hey, we need to get them to use a different computer, so they'll have to buy all new hardware, software, services, and go through the same growing pains they just went through".
Speaking of growing pains, I smh watching different platforms go through our same growing pains, Example: Back in the day, we had primitive security. A 'password' dataset. SYS1.UADS. Now we have RACF, Top Secret, and ACF2. We grew, and developed. The next platform came along, and they had primitive security, and had to grow, and develop. Same goes for operations, backups, automation, etc. So I'm guessing that the TCO comparison done 18 years ago did not include these things. And now the server sprawl that resulted is deemed too wasteful. So they decide to virtualize. Bam! We have VMware and MS Virtual Server, just about 35 years after we had z/VM. Growing pains.
Let's get back to 'acknowledging that it happened'. I think a case study would be a very cool thing to do. The trouble with that is, people have moved on, records are no longer kept, shops don't want to share their failures, or admit they were wrong. So in many ways, I guess we'll never know. But did that first migration off the mainframe have a roadmap? Accurate estimates? I think not.
I also wonder if those who migrated off the mainframe have any idea what they're missing. It's an exciting time for mainframes, with exponential growth in capacity and performance, and some truly remarkable technologies, like out-of-order execution, branch prediction, and transactional execution.
In closing, you can call it server sprawl or server creep; call them pizza boxes or peecees; I'm calling them Tribbles.
Do you get the feeling I watch too much TV? I strenuously object.
Your link today is to my friend Ray Mullins' blog. If you love assembler, you'll love this blog! http://www.catherdersoftware.com/blog.html
Till next time,