John Parkinson is senior vice president and chief technologist for the Americas region of Cap Gemini Ernst & Young U.S. LLC in Chicago. He recently told Computerworld’s Gary H. Anthes why he thinks IT outsourcing and agent-based systems will become ubiquitous, why collaboration software will not, and why data warehouses and operating systems might disappear altogether.
What’s the future of outsourcing?
Ten years out, there will be less than 100 companies in the world that run their own IT. Everyone else will move more and more of their IT into what looks like a utility model, some kind of buy-it-by-the-drink consumption model. There are some things that have to happen for this to catch on big-time. Security and confidentiality of data have to be addressed in a fundamental fashion. But that’s happening.
How will this utility model work?
The drive will be to outsourced shared instances — where there is one copy of the software but 20 companies on it. The vendors have figured out that one-instance company licensing is over because companies won’t pay that kind of money anymore. So you take what SAP does and break it into a series of services, then each company simply pays for the services it consumes. All the big guys are moving their product architectures in that direction, and it will take one to three years to complete that transition.
What’s coming in information security?
Vendors, after a decade of pretending it wasn’t a problem, have finally figured out it’s not enough to put patches out; you’ve got to make them really easy to apply. We’ll see more (automated) enforcement and simpler administration. And there will be role-based security; when I show up and announce myself, what I get to do is dependent on the role I’m currently playing. This becomes essential, because as the complexity of technology increases, and as the range of tasks that knowledge workers do increases, it’s going to be too hard for you to remember who you are supposed to be for the next 10 minutes. I see a five-year horizon to get all the pieces in place.
What are the pieces?
These things will be in the operating system, and there will be security APIs available to any application. There will be standards, like SAML (Security Assertion Markup Language), that applications must conform to or they won’t install into your data center.
What will the arrival of ultradense, very cheap data storage mean?
We are within three to five years of having a (portable) device that could hold everything that you’d ever want. Then you don’t need centralized storage. You could have all business data basically portable, walking around in a huge, billion-node peer-to-peer network.
There’s a strategy called LOCKSS (lots of copies keep stuff safe) for doing backups. The theory is, if there are enough copies in the network, you can always re-create what you want if you lose it. With a billion nodes, and every node storing a couple of terabytes, it would be relatively easy to do away with those specialized data management architectures that we have today.
What’s coming in data mining and analytics?
With federated (distributed) data management, we stop needing data warehouses. We have instead embedded analytics, so applications simply use transaction data on the fly. Over the next two or three years, we’ll see a lot of emphasis away from conventional data warehousing and toward the embedded analytics, toward on-demand business intelligence.
Will agent-based systems based on complexity theory become commonplace? Yes, within five or six years, because we are at the limit of what we can architect reliably in a large, complex enterprise. There are too many moving parts. If you try to design it deterministically, you discover you can’t predict all its behavior. This will (make) obsolete almost all the architecture design knowledge now in use.
How will operating systems evolve?
One scenario is they will go away, shrink and simply become the DNA of agents — much simplified and localized. The other scenario is the opposite — that they will subsume almost everything else and become the service-provisioning platform. The service-oriented guys are going for the latter, and complexity theory and agent guys are going for the former, so we will probably end up with a mixture.
What’s coming in groupware/collaboration?
We are a long way from having a worthwhile collaboration technology. The problem is, the technologies are attempting to make us work in patterns we wouldn’t normally choose. The benchmark for utility is the telephone. After voice, e-mail gives you about 95 percent of the benefit of connection. And e-mail plus instant messaging gives you about 99 percent. The next piece is really hard. And the only reason we are even thinking about it is because we have so perverted e-mail and IM with noise that we are now convincing ourselves there must be a better way.
How might logistics systems evolve?
The problem with ERP systems is that they focus on the enterprise, while the real problem is interenterprise. If you can make things work really well across the players in a supply chain, you can produce huge operating-efficiency improvements, and that’s where agents come in. Supply chain systems will become much more responsive. They will know who’s in the store, what they are likely to buy next, what should be on the trucks in transit, what should be in the distribution centers and what should be ordered — all adjusted in close to real time. We’ll have the technologies to do that within five years.
