New Daedalus

Peter Carbone, Vice President of SOA for Nortel, gave a nice high level talk at the OASIS conference on the challenges facing a company learning to dance in the world of SOA and mash-ups. Nortel, of course, grew up with rigid account control and vertical integration in a regulated environment. As markets for building systems are still characterized by rigid account control and vertical integration, and the power grid is still vertically integrated, regulated, and almost complete account control, there are some useful lessons.

Infrastructure convergence was the enabling and driving change for telecommunications. Provisioning telecommunications was long the most difficult task. Over the last decade, the diverse communication infrastructure converged to a single packet-based infrastructure with resulting dramatic simplification of security and reliability. The questions move from “What low level communications do you need” to “What interactive services do you need?”

This evolution changed how Nortel had to think about and market their services. Before the change, Nortel sold vertically integrated applications that were inflexible. As the core technologies converged, Nortel was forced to decompose advanced services into core functions and then plug them back into the new architecture.

Fortunately, decomposing integrated services into core functions looks a lot like defining a service for service oriented architecture. Fundamental telecommunications functions can now be built into enterprise applications without requiring exotic skills are deep domain knowledge.

Skills-based routing and deployment was one example. Peter discussed a SAP integration with critical system causing expensive downtime, emergency part ordering, and synchronizing communication with an outside expert so that the repair personnel, the piece of equipment, and, via telecommunications and real-time identification of the expert on call, the expert’s telepresence were synchronized.

In a similar vein, he discussed abstracting the GPS function from the cell phone to block access in the security system when the phone was in a forbidden zone. Peter gave many more examples and you can find his slides on the OASIS conference site.

So what can building systems and the power grid learn from this?

Well, the owners expect the systems to just run, and are annoyed when they are expected to learn terms like BACnet or LON (or any other control protocol). We need to decompose advanced services to discover the core functions, from the owner’s and the tenant’s perspective, and present them as interfaces that can be plugged back into the enterprise.

As Peter summed up the C-Level response: “I just spent $100 Million fixing my processes, you had better be compatible.”

Building services that can present themselves as that can interact with SAP, or with PeopleSoft will have an advantage. The services that know how to display themselves on Google Earth will know how to request the nearest technician.

Likewise, Grid requests that present themselves to ERP services will find faster acceptance. Grid requests that describe grid pricing as shapes that can be pinned to Google Earth will enable the enterprise to come up with multi-site responses that may be different from any single site.

No one cares about the old vertical applications. Enterprise interactions are everything.

This is why the Building Service Performance Group at ONTOLOG (just goggle it) is meeting tomorrow.

Everywhere we look, we see more higher-level, almost cognitive functions being incorporated into low-level products. Cameras are internalizing much of the craft of photography. GPS systems are comparing notes with their peers to provide up to the minute routing choices. Cars tune themselves on the fly, adjusting carburetion and suspension in real time to respond to driving style. Systems are becoming autonomous, competition is moving from commodity functions to service, and markets are starting to turn around interactions and integration.

Yesterday’s Wall Street Journal (see link below) described FotoNation and the software that it provides for many brands of digital cameras. I knew about anti-red-eye electronics. I thought it was a neat trick for the camera to automatically focus on faces in the foreground rather than the between them in the center of the view. I was amused at the camera that would alert the picture taker that someone blinked. The camera that delayed the shutter until everyone was smiling was pretty neat.

The ability of a camera to recognize particular faces in the crowd, and make sure that they, if no one else are in focus was different from the rest. Simply take several pictures of your family and friends, and notify the camera. Thereafter if six mothers, standing in the same place each snap a picture at the third grade play, the six cameras will make different decisions and each mother will find her own little Billy in perfect focus.

We now have consumer electronics with complex learning behavior that it applies to its canned pattern recognition tricks. This is customization far beyond the last generation of, say, a car remembering driver preferences for mirror, steering wheel, and seat.

Consumer systems now cover for the amateurish efforts of their operators to produce first class results. Harried amateur photographers get assistance to achieve professional results. Drivers can get performance out of their cars that previously would have required long practice. Trip planning now acquired the knowledge of a local and an instant awareness of traffic conditions.

Building systems face the same issues and are moving in the same direction. Not only are they often operated by amateurs, but the may be maintained by the insufficiently trained—following their installation by the low bidder. Traditionally, systems have been oversized and over-built, to cover these predictable problems. This leaves a lot of energy and operating dollars on the table. The best systems will move instead to make their systems resilient, as are the camera and the car, and self operating.

This will change the tasks asked of control systems, and how they are integrated. Self tuning systems do not need to share low-level details with those far away. Low level protocols will be confined inside autonomous systems, and only higher-level services exposed. These interfaces will be the basis for next generation integration.

Systems will use these newer interfaces to negotiate service provisioning with each other. Although each system should work alone, they should be able to discover resources that each other makes available. Imagine systems advertising their waste heat as a resource, and then the heat source broadcasting when it needs to shut down. These interfaces will be developed as agents; they know their missions, they defend their missions, they act independently.

Integration will come to assume autonomy, for the new interactions will rely on each system doing what it says, and meeting its contracts. Contract-based integration will increase the value of cognitive performance, as they become the only competitive edge in a world of commodity electronics and unpredictable installations.

And systems that expect to be told what to do, rather than simply meeting their contracts? Well, as now, no one will want to do business with such agents.

As we move from control-based interactions to agent-based interactions between systems, the best systems will use surrounding networks to develop a situational awareness that will set it above its competitors. We won’t be aware of these activities; they will occur without us even noticing. This situation awareness will fall into two categories: awareness of peers and awareness of surroundings. Surprisingly, this awareness will be part of competition on function at the same price-points as systems without awareness.

This morning, Amazon began selling the second, re-tooled version of the Dash Express. The Dash Express is an in-car GPS that does roughly what other in-car GPSs do, and at a similar price. What’s different about the Dash Express is that it is continually attached to the internet. Some out of the box thinking leverages the capability to provide new services.

The obvious stuff is that the Dash Express can keep its traditional GPS functions up-to-date. Was a street renamed or a road completed this week? Changes to Dash maps are automatically downloaded. That new restaurant that you just read a review of? Dash can find it on the internet and guide you there

You can also use the internet yourself as you drive along. You can browse using the Dash Express, and then use Dash to guide you there. Please, just pull off before you do this in traffic. . Looking up places of interest on your home computer? You can send them to your Dash. Dash Express is aware of the changes in its environment and keeps itself up-to-date.

But this is not why I am writing about the Dash Express.

Each Dash Express also transmits information to the internet. It can check the current or recent road speed on each alternate route and choose the one for you accordingly.The more Dashes are on the road in your town, the better this information is. For the first time, we are seeing network effects in GPS performance.

(Network effect refers to markets in which the value of each item increases the more that are on the market. The classical example is the Fax. The first Fax was not worth much until the second fax was plugged in. Once most office had a Fax, you almost could not afford to not have one.)

It is interesting to speculate if we will see emergent behaviors in city traffic in town where there is a sufficient density of installed Dashes. What effects will a stadium parking lot of tailgaters with their engines on have on Dash calculations? Will traffic start to pulse as Dash advices crowds to not take that onramp? Will people watch for installed Dashes as they watch for radar detectors now. Taking a free ride on the technology?

Soon agents in each house will be able to receive “prices to devices”. If I am right, they will aggregate device information through a house agent. Will house agents negotiate with other houses to aggregate larger blocks of demand to sell? Will they ready our Google Calendars and be extra aggressive when we are away from home?

It’s hard to know. But situation awareness leveraging network effects at the same price point does raise a lot of possibilities.

Power grid reliability, human heat pumps, and data centers as energy resources – what is the common thread? All of these rely on being to get above the details of the systems to see interrelationships between the systems. This approach requires systems to compete on delivering of service, rather than focusing on process. Systems that provide a similar service, albeit with fundamentally different internal processes, must be swappable.

We must move beyond protocol interoperability to informational interoperability.

In engineered systems, interoperability usually means “we can get some signal of some kind between systems”. That signal is data oriented, meaning it is a raw fact that is neither actionable nor useful on its own. Someone with deep domain knowledge program the interactions around those facts. This leads to over-integration between systems.

Informational interoperability raises the bar, by allowing systems to compete on performance and service. Data is not information; often too much data can hide information. Only when facts from the underlying process are assembled into patterns that have meaning and can influence action does data rise to the level of information.

If you have two or more systems that can both consume and produce the same information interface, then those systems are informationally interoperable. If several external systems share the same informational interface to the local system while performing different services, then the local systems interface is reusable.

If I am performing an energy intensive task such as intake reheating, it matters little if my heat source is electric coils, a central steam plant, a solar thermal collector, or the data center downstairs. Each has a cost (which may even be negative), each has a quality, and each has performance characteristics. Systems with informational interfaces can select or which thermal source to use, either at design time or on the fly. Such systems would not need to know any details about the internal operations of their design source.

The best system interactions are built using reusable informational interfaces. The most accepted and best understood reusable informational interface is money. Money provides actionable information about scarcity and value. Monetary interfaces are highly re-useable and interoperable.

Bad systems hide information about performance, scarcity, and value; good systems expose such information in ways that allow innovators to take advantage of this information. Let the systems use whatever low-level protocols they want internally. On the outside, we need information interoperability.

Last week five power plants in Florida went off line following a problem in a substation. Active discussion ensued in the blogosphere. One of the first headlines was “Terrorist attack not suspected in plant failures” One of the first comments I saw was on the lines of “Great. Now the IT guys will all come on-line and tell us how we should have done it”.

I’m not going to do that. I have nothing useful to say on the design of any power plants, let alone nuclear plants. All systems were performing as designed. In the belts and suspenders world of nuclear plants, the entire grid is one of the redundant power sources for the cooling systems. The plants were supposed to shut down one of the safety systems lost redundancy. What we saw in Florida was carefully designed systems doing what they were designed to do. If every one of my ideas were fully implemented in the grid, in the building systems, and in building design, these plants would, and should, still have shut down.

If the grid as a whole were re-built as interoperable services with economic interfaces (prices), the blackouts in Florida would not have been as far reaching in their effects. The service oriented grid will enable an ecosystem of local reliability and storage. That ecosystem will support innovation and technology diversity at the distribution and building level. (Note: In power, transmission refers to the long distance transport of energy, the high voltage towers marching to the horizon; distribution refers the lower voltage movement of power around neighborhoods). That market will create islands of reliability wherever it is worthwhile.

The key element is informational interoperability. In engineered systems, interoperability usually means “we can get some signal of some kind between systems”. That signal is data oriented, meaning it is a raw fact that is neither actionable nor useful on its own. Someone with deep domain knowledge program the interactions around those facts. This leads to over-integration between systems.

Very good systematic thinkers tend to extend their systems beyond the domain in which they are skilled. Power engineers tend to build a single giant robot covering continent-sized territories. Faced with the diversity forced upon it by scale, this robot becomes more and more brittle. The only response within the paradigm is for the engineer to become more and more controlling, which ameliorates the systems but makes the long-term problem worse.

Bad systems interfaces hide information about scarcity and value; good systems expose such information. Power systems hide information about scarcity, value, and reliability in systems without interfaces. Utility regulators simplify system interfaces to support historical practice rather than innovation.

The best system interactions are defined around reusable informational interfaces. The most accepted and best understood reusable informational interface is money. Money provides actionable information about scarcity and value. Monetary interfaces are highly re-useable and interoperable.

If we had good informationally interoperable interfaces including a substantial monetary component between each system in the power grid, the plants at Turkey Point would still have shut down. They are well designed systems engineered for safety and long-term reliability. What would be different is that their customers would not rely solely on the fragile power robot. What would change are the local markets in reliability that would spring up. Local markets would let new classes of innovators seek profits in providing new value.