AutomatedBuildings.com Column - Irresistible Data Forces and Unmovable Building-Based Systems

Last month’s magazine developed Ken Sinclair’s vision described of Humanistic Digital Inclusion. This compelling vision will require agents on the Internet of Things (including buildings) that are able to anticipate the needs and wants of people and marshal the services to provide them. This will create a premium around local knowledge as each building, and each environment has hard local requirements. For example, a building that has been allowed to grow too warm in the summer may be unable to cool on demand to the target temperature required by the human-centered agent. Agents for the human and for the building will need to negotiate to come up with the best outcomes.

It is a mistake to think that our current application paradigms, always driven by the last big problem, will be how this comes together. Today, we imagine all that all applications must be cloud-based because we live on the internet and get so much wonderful content for free. Somehow real monetization will flow from things we learn across all platforms, just as today, real wealth falls out of web tracking and advertising. Others hope they will find value in huge masses of data, just as the cheap gene sequencers (23&Me et al.) give away services as a means to someday own personal actuarial data to sell to financial services companies.

This month, the first tangible wave of the global revulsion against this model rolls out, and websites everywhere are asking for permission to track under the new rules. In the US, the data harvesters such as FaceBook and Cambridge Analytica have been chastised by many who while complacent when data harvesting supported purposes they approved of, are now horrified that the same data and same techniques can also support purposes they disagree with.

In buildings and physical infrastructure, the new value propositions in IoT-based services are most valuable where they cannot be used with today’s techniques. Tracking and sharing of Fit-Bit data have put at risk military and security operations worldwide. Highly structured operations in critical pharmaceutical manufacturing facilities become easier targets for disruption as operational data leaks into the world. Easy remote access becomes more and more security risk. Smart transactive microgrids can create needed resilience for high-value facilities, but if cloud-based, are poor fits for facilities that wish to cloak their electrical use to occlude power signatures.

For the most critical resources, the communications channel itself becomes the Secure networks have been compromised as individual smart lightbulbs and thermostats become targets to enter entire secure networks at will. Loss or interruption of connectivity, whether through equipment failure or through line interruption, whether from natural causes or from enemy action, becomes unacceptable as the criticality of the systems increases.

The work of IT Architect Pat Helland has long provided useful models for thinking of this sort of change. His landmark 2004 article “Metropolis” on real-world technology evolution should be read by anyone contemplating large-scale multi-system multi-vendor build-outs. He outlined the challenges of widely distributed data and diverse distributed processing in the talk “The Irresistible Forces meet the Immoveable Objects” in 2007, which addressed the issues only now coming to the fore. More recently, Helland has written on the importance of immutable data for new systems, and even that data immutability is more important than data normalization.

Data Immutability can be understood as the principle that once written, data is never changed. If the data is signed, the signature is verified, and the signature cannot be changed. By analogy, data immutability is the accounting principle that ledger entries are written in ink. Entries can be backed out with a new entry, but never deleted.

Nowhere is data immutability more important than when AI (Artificial Intelligence) is applied to the IoT. Operational AI finds patterns in how things work and then uses those patterns to make things work better. Even the most secure AI Agents can be co-opted by feeding them bad data that will create incorrect patterns. Incorrect patterns can induce behaviors whose effects can range from merely disruptive to financially expensive to imminent dangers to health and safety.

Cloud-based systems create data immutability by managing data on traditional databases that are managed in the cloud. Professionals manage data back-up, and whoever can see that data can see all data. Loss of access to the data is as harmful as loss of connectivity to the cloud.

With traditional databases, local data storage becomes a risk. Bad data can be introduced locally into the database using common tools. Local databases must be backed up and managed just as professionally as the ones in the cloud are.

Cryptoassets is a term that encompasses the techniques used in cryptocurrencies such as BitCoin and Ethereum. Cryptoassets use specialized hashing algorithms to create distributed immutable databases. When these databases have a local scope, with no need for universal access and scale, these databases can be built using small, inexpensive systems such as the Raspberry Pi. Just as a disk RAID set can lose a member and still retrieve all data, so can a hashed distributed database a member and still have the entire dataset.

New systems can be introduced to and begin participating in such a system. Other agents, with appropriate authorization, can be permitted to read all or part of this immutable database. The system can share access to operational data with third parties, such as a utility or accrediting agency, that pays for access and trusts this database does not need to upload data constantly. Operational AI-based Agents can interact with such databases with confidence.

For years, I have described the imminent arrival of small system-based AI in smart buildings. Last summer, Microsoft released an open-source GitHub of AI for small systems including the Raspberry Pi. Last month, Alper Üzmezler revealed his SandStar platform for autonomous building control components. Couple these with an immutable database to support the AI and the components for real learning-based building systems are in place.

Such AI systems will be needed to fully embrace Ken’s vision of Humanistic Digital Inclusion. The larger database can learn about individuals within a building and pass hints to each system. Those systems can store new information about preferences in the building-wide immutable database. The remaining question is how one building can tell a peer what the human likes or wants. That will be an interesting set of standards to watch develop.

Events	Want Ads
Our Sponsors	Resources