Dirk Slama: Some argue that the cloud has been a massive effort in centralization, while the IoT is now exactly the opposite – a massive decentralization, with logic and data scattered across billions of distributed devices. So is the IoT killing the cloud?

Peter Coffee:  I don’t agree that the cloud is a massive effort in centralization. Yes, there has been a certain amount of centralization, people moving from local word processing to collaborative document management, for example. This is relevant, but it is not the most important aspect of the cloud. The most important aspect of the cloud is that it allows you to stop thinking about IT in terms of physical assets that have to be acquired and managed. The cloud is about building application services by orchestrating APIs (Application Programming Interfaces), pulling data, and leveraging machine power in a highly location-independent manner. The way we work is completely decentralized, because everything is now abstracted and distributed in the cloud. Previously, taking things outside of data centers was difficult. Cloud is diffusion. So from that point of view, the IoT is only the next step in the ongoing decentralization effort!

Dirk Slama: What are the key drivers in this?

Peter Coffee: In the past, computers used to be scarce. Now they are ubiquitous. Then connectivity was scarce. Now connectivity is ubiquitous. What´s scarce today is trustworthiness of data and of the partners with whom you are sharing data. What is relevant today is added value, or the potential you can leverage by securely operating on that data. This is the area of the enterprise cloud where we offer services to large businesses and government organizations, by leveraging a global pool of capabilities – placed where the availability of power, space, and network connectivity permit.

Dirk Slama: So how do you address the issue of trustworthiness?

Peter Coffee: It really starts at the core – data is born naked and defenseless. There is nothing that protects it from the world around it, proves its authenticity, etc. We built a database where you cannot find a single data entity that does not have full traceability. Each entity is protected, privilege managed, and fully traceable. Having this kind of metadata (3) available for each entity is totally different from traditional databases. From an IoT point of view, this applies both to the core asset master data, as well as the context data, such as the customer related to the asset. And this is the foundation for our multi-tenancy capabilities (5). The ability to securely run many thousands of tenants on the same system, all securely sandboxed, is really the key differentiator of a cloud solution like ours. Yes, we are also a great CRM and Service Management Platform, but multi-tenancy is what enables this in the cloud, and provides the technical foundation for trustworthy data management.

Dirk Slama: Putting the data from assets and devices in the right context is also important?

Peter Coffee:  Yes. Andrew Rosenthal from Jawbone [a wearable devices company] once said that getting the data from the devices is actually not that interesting. What matters is putting it in context. Data only starts to become meaningful if you start putting a story around it. Counting your steps is not important. Being able to tell you how many calories you have burned today compared to your wife – now that is interesting!

Dirk Slama: So what is Salesforce’s IoT strategy?

Peter Coffee: Salesforce has never been a mixed strategy company. We have always been a cloud service provider, not a software company. So there is no need to discuss our transition. We were born “cloudy.” We have recently been listed in the top 5 of MIT Technology Review`s “50 Smartest Companies,” together with Illumina, Tesla Motors, Google, and Samsung. The reason MIT gave for our ranking was that our “tools will be crucial in helping companies incorporate new data from the Internet of things” [MI1]. The IoT will generate an enormous amount of multi-dimensional, textured data. For us, this is an incredibly rich opportunity to help our customers find out what their customers would find delightful before the customer can even articulate this themselves.

And we are obviously not a device company. Take Apple with its upcoming Apple Watch product. Apple Watch will trigger an ecosystem of apps that will each generate data. This is of interest to us. Or take Scania. They recently launched the Scania Watch, which provides truck drivers with information like transport metrics, fuel data, and driver support scoring [SC1]. This is not only brilliant in terms of brand building and customer loyalty, it is also a great example of the potential of wearable technology. Providing the right tools to allow integration with any kind of wearable is a key part of our IoT strategy. We recently announced Salesforce Wear Developer Pack, a collection of tools that let you design and build wearable apps that connect to the Salesforce1 Platform.

Dirk Slama: And what about the Industrial IoT?

Peter Coffee:  Yes, also extremely important. But again, for us this comes down to putting data from assets and devices in context and then adding services. I mean, collecting data, analyzing it, summarizing and taking action is not a new concept. I worked in the oil and gas industry and saw the first digital control rooms in refineries emerge in the 1980s. You would be amazed how much added value you can create by adding the right context to what is actually very little data to begin with. Take an oil field with hundreds of pumps, spread tens of miles wide. The most useful thing to know is whether the pump is actually pumping or not, so that you can give the field workers directions on where they need to go to do something useful. Or take repairing a wind turbine: this requires climbing up a 100-foot tower. So it would be nice to climb up with the right tools and instructions in the first place, instead of having to climb up twice.

This is exactly the role that cloud-based services like ours can play in the IoT: bringing data from the IoT into processes and putting it into context (4). So if the telemetry data is telling us that the wind turbine needs fixing, we can find an available service technician, tell him what equipment to take with him, and what vehicle to use from the repair truck fleet. Ideally, this would actually be done even before anything breaks down, because a breakdown is a customer experience that can never again be repaired after it’s happened. So developing the capability to predict maintenance situations will also become important.

Dirk Slama: So what kinds of feature does a cloud application platform have to support to enable these kinds of IoT scenario?

Peter Coffee:  I think this is all about the cloud engine APIs (6). Many cloud applications have traditionally been built around web-based UIs. Later, mobile capabilities were added. However, in the IoT, in many cases the first line of contact will not be a human user, but a machine or some other kind of asset. For this, it is important that the data coming from the asset can be processed automatically – and this requires direct access to the cloud engine APIs. Salesforce has completed the migration of all cloud applications to open APIs in 2013. This gives us a great deal of flexibility for developing mobile and web UIs that aggregate different types of application (7). And I also think this was a milestone for us in terms of IoT preparedness.

Dirk Slama: What about the technology required for data acquisition and device control?

Peter Coffee:  I think it is unlikely that we will be in that market place. We have many partners in our ecosystem that provide these kinds of capabilities, such as Etherios [recently acquired by Digi]. I also think it is important to realize that the IoT will not be implemented on a green-field environment. Similar to our situation when we started in an existing ERP market, this is about adding value, not replacement. But of course, new cloud solutions can also add value in the integration space. Take, for example, cloud-based SCADA. I think in the next decade we will see a lot of growth in boutique manufacturing, factories that are able to produce small amounts of highly customized goods. They would not be able to afford a full-blown, self-administered on-site SCADA. But with cloud-based SCADA, connected to services such as ours, this becomes a different ball game. And we are quite happy with this. At the end of the day, we are not an industrial automation company: our job is to add value to data, not to originate it.

Dirk Slama: So the IoT integration layer will be more like a Platform-as-a-Service (2) (PaaS) model that integrates with a Software-as-a-Service (1) (SaaS) model such as yours?

Peter Coffee:  Yes. Think of it as the “last mile” problem. Somebody needs to address this, but it will always be more low-level, integration stuff, including protocol bridging, data stream analysis, etc. Except that it will become much easier to address the last mile in the future because of the advances in technology. In the past, the local carrier owned the cable into the basement of your house. Nowadays it has become very easy to attach a box to a street lamp pole to provide the required bandwidth to the houses in the vicinity. Of course for trucks and container ships this is still a different story. So yes, PaaS in this context is relevant from an integration perspective.

Dirk Slama: What kind of logic do you see residing on the asset in the future, and how much in the cloud?

Peter Coffee:  I like to compare this to how the human body works. The sensory neurons do not exclusively pass information to the brain. Local spinal motor neurons can trigger reflex actions almost without delay. So if you touch something hot, your hand moves away automatically, while your brain starts putting the information into context and thinking about finding an emergency exit route. In the IoT, we see poppy-seed-sized microchips that can do amazing things when deployed on an asset. But putting things in context and making strategic decisions is something I will always see in the cloud.

Dirk Slama: So physical and technical boundaries (9) remain important?

Peter Coffee:  It is extremely important to understand the latency and bandwidth restrictions in your specific environment to make decisions about data and logic distribution.

Dirk Slama: Time sensitivity is also often cited as a big issue. It seems unlikely that an application like controlling a high-speed manufacturing robot with hard real-time requirements can be done out of a cloud environment. Or take the CERN case study from this book, which describes a massive, multi-tiered analog/digital data conversion, data collection, distribution, and analytics application. Another unlikely cloud candidate?

Peter Coffee: This might be true at the moment. But I don`t think the accomplishments of the cloud should be viewed in terms of physical distances or economies of scale. Don`t forget that we are living in a world of constant technical evolution. The existing boundaries are moved, every day (10)! In this world, the cloud should always be the goal. So instead of the silo being the default, the cloud should be the default.

At the point of data origination you can first decide if you need to process it locally, or make it available over a bus, or a local network in a building, or globally to all devices of that type – at which point you are on the Internet and in the cloud. If you originate the data in a form that it is accurate and trustworthy, the best way to put it into context is in the cloud.

Also, think of emerging concepts like micro clouds (8), which can be deployed locally. In your discussion about the connected vehicle, you talk about the future, sandboxed application platform for the car. The foundation for this will be nothing other than a local micro cloud, connected to the Internet cloud.

Dirk Slama: So what is your advice for the IoT project manager and their solution architect who have to decide on a cloud strategy today?

Peter Coffee: Understand today`s technical boundaries. But design your system in such a way that it anticipates and allows the continued movement of these boundaries instead of engraving the current boundaries into the design. The question of “what is cloud and what not” is not an answerable question: micro clouds, personal clouds, clouds of communities of devices are emerging. Think of them all with the same cloud perspective. Don´t build silos, but clouds instead.  Allow for departures from the original strategy, but anticipate that, over time, the need for change will disappear.