26 December 2012

The charm of history and its enigmatic
lesson consist in the fact that, from
age to age, nothing changes and yet
everything is completely different. 
   —Aldous Huxley

I just read a report in the Motley Fool's website about the Netflix outage over the holidays. According to the report, the streaming video service had a catastrophic failure on Christmas Eve. At fault was Amazon's Web Services unit, which Netflix uses as the storage and delivery vehicle for its videos. The outage lasted more than seven hours and affected Netflix streaming video customers in North and South America.

Somebody should ask all the people who hoped to enjoy an evening of holiday movies with family and friends how much they enjoyed the change in their plans. Ask them while you're at it how they feel about centralized controls over their viewing options.

Thirty years ago, I worked for a software company that created a 4th generation relational database management system (RDBMS). Anyone who works with databases will tell you that they suck up resources like nobody's business. Once the data is loaded into the database, the sheer volume of information requires extraordinary amounts of storage. Back then, in the days before personal computers, the cost of storage was expensive, usually prohibitively so. The data center at our little company occupied 8,000 square feet of prime real estate. We filled the space with Prime, Data General, IBM, Honeywell, Harris and DEC mini-mainframes, system printers, switches, routers, miles of cables and wires, and banks of gigantic tape decks. The amount of heat such a collection generated was considerable, so there were matching Liebert cooling units stationed at either end of the center. The monthly utility bills were always thousands of dollars.

But somewhere along the line, a genius in IBM's accounting office concocted a great plan for reducing the cost of computing for their customers. Instead of sending equipment and support out to customer sites, why not have the customers subscribe to an offering that would give them equal access to the equipment and services they needed, without actually having to lease or own them? Think of the savings if customers didn't have to support a full data center -- no more expensive computer administrators required, no more expensive machines that needed arctic cooling and unending supplies of electricity, no more paying for idle time and unused CPU cycles -- truly, a CFO's wet dream. They called it "time-sharing."

IBM put together gigantic data centers and marketed them to small and medium-sized businesses all over the country. My company signed up; we had three big IBM minis in our data center. Our CEO was anxious to jump on board as fast as he could, anxious to begin realizing the promised savings. If the experience was as good as what they promised, he'd do the same for the other computers in our data center and shut our operation down.

Six months later, we all felt he'd jumped aboard the Titanic, just before it met up with the iceberg.

Transitioning the data from our data center to IBM's was a Sisyphean chore. After two full months of churning, in desperation, we made the decision to shut down the IBM computers in our data center and physically ship them to IBM, where their administrators could migrate the data from our machines to theirs.

Once the data was transferred, IBM sent trainers to our company to familiarize us with our new service. After a week, three senior programmers resigned, declaring they'd rather be unemployed than tethered to IBM's rigid, hide-bound practices. One thing I know for sure is that creative people hate being constrained, and at that time, IBM was all about operating procedures, rules, and limitations.

Then came the actual implementation. In order to use the IBM service, we discovered we weren't going to be able to access the services from our own offices. We had to travel to Needham, where IBM had created a "service center" in partnership with a local company only too glad to collect rent from them for the privilege of letting IBM's customers sit in assigned tiny cubicles for pre-specified times of day. We had to be verified as legitimate users, authenticated as having the right to access our own data, and monitored completely to ensure that we were performing actual company-sponsored work, rather than simply stealing IBM's vast computing resources.

The indignities didn't end there, however. We had to pay extra for hardcopy of what we were working on, but we couldn't have printouts unless they were pre-approved by both our company's management and IBM's data management administrators. Because our CEO was really focused on savings, he'd arranged for us to have "non-prime" hours. This meant we could have access to our data services between midnight and seven a.m.

I assure you, technical writers, release coordinators and support technicians do not appreciate being called to duty on the graveyard shift. We lost a bunch of them. Needless to say, our IBM-based customers weren't thrilled to learn that they couldn't get answers to questions or issues in anything approximating a timely fashion. We lost some of them, too.

Before the grand experiment ended, we'd said goodbye to several valuable long-time employees and a few disgruntled customers who headed off to our competitor (at that time, another little RDBMS company called Oracle). Our relationship with IBM was in snarling tatters, a situation made worse when we had to threaten them with legal action if they didn't transfer our data back onto our leased/owned machines and ship everything back to us.

We ended up right where we started, but poorer and a lot wiser. Our CEO admitted it was the worst tactical error he'd made in his then-lengthy career. He predicted that a time would come when people would recognize the limitations of centralized computing and demand full control over their data and resources. He said that nobody in his right mind would want mission-critical applications and data on remote servers where security and availability could not be guaranteed.

He predicted, rightly, the advent of personal computing.

But here we are, all these years later, and "time-sharing" has transformed into "the cloud." It's all tarted up with new terminology, but essentially it's the same as it ever was.

And now all the Netflix customers have gotten to experience the downside and the emptiness of the "uptime" promises first-hand. Welcome to the wonderland of cloud computing, folks.

Excuse me. I'm late for the only "time-share" I like, which is a seat in the theater where I'll watch a great film and not worry about whether it'll stop in the middle.

No comments: