New classes of information are measured by intricacy, not size

Abstract image of blue and white lines

Being a network person, I thrive on the numbers. Folk like me continue to worry about performance, because we're unlikely to see a slow network that someone is entirely happy with; it's far more likely that we see a huge volume of traffic being squeezed into a tiny network lead, or over a slow circuit. So when I was invited to attend an award ceremony at software giant SAP's global headquarters, I jumped at the chance.

What's that got to do with esoteric numbers? Quite a bit, actually. For one thing, the awards were being given to entire teams of people in SAP's world-spanning customer base. Getting a person through a qualification, or scoring 100% in a test, is hard enough when it comes to vocational qualifications. Getting ten people from the local government of Cape Town to the finals of the SAP Excellence awards is a whole different level of effort. Quite apart from the implicit project budget that can cope with so many people being invited to a small town outside of Frankfurt, from a small town 10,000 miles away at the other end of two continents, I was intrigued: how were they evaluating these contestants? An initial fear that I might be the journalistic window-dressing on a company junket was blown away by a few instant impressions.

The first was from before the ceremony even kicked off. While waiting around, I went through SAP's tour of its digital museum. Some cross-language hilarity ensued, because the lady doing the tour was in her early 20s and had to have it pointed out to her that, quite possibly, I'd used most of the devices and objects in the display in anger and might know more about them than she did.

Actually, she wasn't all that bad. It was bizarre to see that SAP believe people stopped using punched cards in 1980, just one year before I started in an IT job, and despite a little speech she had about how the punched-cards are for "eight-bit computing", she didn't know what EBCDIC was or why it was useful for punched-card storage (see en.wikipedia.org/wiki/EBCDIC). Far more sobering for me was to see the teardrop-shaped, first-generation USB storage key, in its own display stand and annotated as being a laughably small 8MB. That's an "M", not a "G": I clearly remember dragging a client to a shop in Tottenham Court Road in the early 2000s to buy several such devices as the instant solution to not having to lug home huge laptops on the tube every day.

Having established that I was only a few years behind SAP's founders in the IT fossil stakes, we moved on to the awards themselves. My question about appraising the contestants was answered almost straight away: probably 10% of the people at the ceremony were actually judges and sponsors, in the style of the singing coaches on a talent show. They hadn't simply taken a quick look at a shortlist of projects and ticked off the one they thought most deserving: in this kind of software product, it isn't over until the fat lady sings. Continual contact, getting to know the business and the implementation team, realising that a coffee morning is as important as a server upgrade: this is a different way of measuring whether software is doing the right job for people.

As I found when I had a sit-down with some lovely ladies from Tata Steel. Neither they nor I knew at that time that they'd end up as the award winners of the day. What I wasn't expecting was a neat pair of my misapprehensions being blown apart in very short order.

The first came from Thomas Otter, owner of the system design of the product within SAP that Tata Steel is using. He was just my kind of interviewee: knew his product inside and out, without being burdened by that awful marketing person's need to make every answer sound positive. We had a terse engagement while I puzzled about how to relate ERP industrial stock and process monitoring software to a steel plant, where all parts of the manufacturing process can be counted on a safety-gloved mitt. Iron ore, electricity, big equipment in; steel, out. I always think of ERP as being about thousands of parts, hundreds of types of screw the sort of software that does best in a factory that makes washing machines. How could steel-making require big data, cloud service-based, high-availability software products?

Both Thomas and the Tata Steel ladies had the answers. The most important thing to know about the guy in the picture who stands a few yards from liquid metal at 2,500C is that he has the right qualifications, the right experience, the right safety gear, the right training and the right insurance. Keeping track of that isn't about huge quantities of data or about the pot-of-gold promise of emergent properties from bland information: it's about having an enormously flexible database model, and being able to keep information up to date without subjecting it to choke-points around trusted operators, mother hens, curators, or whatever you want to call them. To make that kind of data work for you, said Thomas, they found they needed 150 different types of external connector.

This isn't a wry reflection on the physical, outmoded formats in their computing museum. It's an appraisal of the immense variety of APIs now dominating business computing, each one allowing one piece of software to interrogate another. Even though the average steel plant might have a workforce whose entire database of skills, address, job history, performance and so on might fit on a microSD card in one person's smartphone, that's entirely the wrong place to be keeping it. When the query you run on your HR system could be a matter of life or death, because it tells you whether someone should be allowed to wander about with a ladle of liquid steel in a factory full of stamping presses, you don't want it to be inhibited by poor or convoluted connections in any way.

This is when the penny dropped. I'm sure everyone thinks of databases as being pretty boring things; interest is added only by the analysis process, or by the volume of data being beyond human comprehension. But now there are other classes of information altogether, information that's really only used by large teams of people. This isn't measured by size, but by intricacy: in SAP's case, that 150 connection count is a measure of performance that doesn't relate to speed, but to quality of development, and the degree of market overview enjoyed by the designers.

I came away impressed by the understanding shown by these customers of the products with which they were engaged. Very often in networking, any attempt to show how a more fully featured product would make things better is met with the "oh, you computer people" dismissive reply. It was very cheering to see that not everybody thinks this way.