This is not an aircraft

This statement can seems either weird or provocative, but, after years in the avionic and transport industry, I gathered an important knowledge on processes, interactions and business specifics of this area. However, when it is about thinking about digital, I need to prevent my mind to follow its normal routine and come back to the basics.

In the 80’s, an IT administrator was separating the part of its network that was in his primary facilities (in its LAN – Local-Area Network) and the secondary facilities attached non permanently to the network with a specific connection (often operated with a loudly slow modem).
Year after year, with the expansion of the virtualization and cloud techniques, we moved away from many of those challenges, grouping computers into data-centers, availability zones and cloud regions. Our device seamless jump from a LAN to a WAN (Wireless Access Network) or from cellular to Wifi, taking benefits from strategies of cost saving and search of the most reliable or efficient communication pipeline.

To grasp the right ways to apply pattern of the digital world, it is important to separate the function (like making a data synthesis), the location (like the airport) and the physical object (a computer or data center).

What does this general statement mean in practice ? When we think about a specific process, like the aircraft connecting to the gate to be able to load or unload data information, it is important to remember that this is a computer or a group of computers that are connecting to each others. This statement was at the origin of this famous over-simplified statement : “There is no cloud, just someone else computer”
With this first focus, we look now at the entire Aircraft/Airport/cloud/PED objects as just a set of data nodes that have permanent or temporary links, and attributes that are related to their management or physical reality.

Let’s look at some examples.

From a general cloud provider set of datacenters, we rent several virtual machines that we combine in group of devices that we can consider as virtual data centers.
As this virtual data center operate in a cloud provider large infrastructure, the virtual data center has the ability to grow and shrink based on our need. Each machine that is failing will be replaced by another that will continue or restart the transaction

At the airport, several physical devices will be combine into a physical data-center limited by the ability of the airport to provide physical spaces for those needs (and invest in performant devices). From this physical footprint, we will have the ability to create again a virtual data-center that can grow and shrink but this time, with a maximum limit related to the airport infrastructure.

In the aircraft, a set of devices are installed in the cabin and at the seat. From there, we often create a virtual cabin data center with several devices in the cabin, and a virtual group of devices at each seat. Due to the power supply management of the aircraft and other regulations, the virtual data center will not have the ability to be highly available and its connections capabilities will be affected by the flight phases or physical location of the aircraft.
At the seat, the specificity of each seat device can prevent to use any strategy of redundancy or balancing of services from one device to another.

So, do we need to move back to a static view of the infrastructure ?

A few years ago, when we introduce the notion of immutable infrastructure, the highlight was to enforce the separation between function (like video streaming) and the infrastructure that support it. Moving from a “personalized” view of the system where each system component has its own name and specificity (the pet) to a paradigm where each machine is differentiated by its capabilities (like important or small hard disk) and managed as a anonymous resource (the cattle).
Each evolution of the installation is performed by creating a full brand new server and destroying the previous one (instead of trying to update it)
With other strategies like resource discovery and load balancing, we evolved toward the real ability to have a dynamic and auto-healed infrastructure.

Loosing this capability, would restrict the overall set of strategies that we can apply on our solution to be cost efficient and limits our ability to deliver high performance and high available services.

Create and use metadata on the infrastructure

Each pool of resources (virtual datacenter) need to be characterized by several attributes that will allow to apply appropriate strategies. Each of those attributes are often specific to the device, but it is beneficial to define them at the group level to optimize the speed of the strategies application

Let’s list some of them

Power supply.
This characteristic will highly impact the ability of the device to manage its resilience (backup of its state and data). Many devices now provide the ability to detect the lost of power and to have a minimum amount of time to perform some hibernating operations.
As a reminder however, applying 12 factors principles and micro-service or stateless architecture allows to have limited need of those operations.
However, a device that have limited period of power supply availability is not suitable for hosting background processes that need for example either data resilience or long computation time (like data correlation algorithms)

Extensibility
A few years ago, when working on inserting virtualization technology in industry production lines, the challenge was to find a smart way to combine physical needs (like measure an analog signal) and virtual needs (abstracting the physical machine to create pool of resources).
Defining the extensibility of the device allow to understand if the device can accept or access to physical extensions (like a seat belt sensor).
In average, virtualized resources of cloud providers or physical data-centers do not allow physical extensions as it contradict the virtualization hypervisor architecture used in those systems

Geo-location
The geo-location is the information that will allow to define appropriate strategies of proximity to a user, data movement ability or other physical contraints related to the location.
Legal country policies will, for example, limit the ability to transfer data from one country to another.
As a warning however, the geographical attributes needs to be limited to the information that makes a difference in the appropriate management of the device, otherwise we can have a tendency to mix digital distance (network latency) and physical distance.
For example, if a computer located on the second floor is virtualized as part of a data-center, its physical location shall have no or little impact in its usage

Inter-dependencies
Even if virtualized, some devices can have inter-dependencies (for example to the same power supply) that are important for strategies like high availability or ramp-up processes.
The inter-dependencies are often specified as sub-group in the virtual data-center (or region)
One example of wrong behavior created by ignorance of inter-dependencies is the “Domino effect” in data-centers that will cause a data-center to go down when all the users are connecting in the morning and overloading one by one the power-supply and computing capacity of each blade of the data-center

Capacity of processing or storage
The local capacity of the device is a choice criteria for appropriate usage of the devices (and often management of its associated cost).
To have efficient resource usage, the system shall manage two important goals : select a capacity that is the more closed to its need (measure of average and pick usage), and having enough spare to have the appropriate level of resilience and extensibility in the service availability (for example, load data in memory for processing).
In average, as human are bad as selecting between too many choices, the cloud provider groups similar devices into class of devices (small, medium, large …)

Communication capability
Some devices (like a tablet) will have intermittent communication capabilities based on their location, state or physical connections.
For example, many aircraft SATCOM modems can only be used above 10000 feets
Some other devices will have limited total capability of transmission due to the protocol (like text message) or the technology (old fashion modem or bird communication :-))

Mobility
Some devices (like tablet or the SATCOM modem) will have direct mobility capacity; which mean connect to several data points based on their location.
Appropriate characterization of the mobility will allow to manage the strategy of the device to create and maintain a stable connection with other devices

Apply the digital strategies

This list of attributes is not exhaustive, as several other attributes can be defined to enhance the global resource management and data communication strategies of the solution (like power autonomy from internal battery).

However, the point is to understand that the environment that surrounds the device or the virtual data-center shall be captured only if it affects the device usage capacity of strategy.

To understand the range of this important approach, let’s look at one example.

When providing Internet access on-board to the passenger, several environments will be involved

• The Wifi network inside the cabin between the device and the WAP (Wireless Access Point)
• The network between the Wifi access point and the SATCOM modem
• The network created between the SATCOM modem of the aircraft and the SATCOM satellite
• The connection from the SATCOM satellite to the ground station
• The connection of the ground station to the SATCOM data concentrator
• The network bridge that connect the data concentrator to the Internet service provider.
• The Internet network

Despite all those differences (by the way amazingly interesting), the success or failure of the user navigation shall be abstracted as a data path involving several nodes (WAP, SATCOM modem, satellite …)
This abstraction shall be the only model used for strategy of Qos (quality of service), troubleshooting (non availability of the Internet site) or data optimization (prediction of connection lost or limitation of traffic capacity)

With this approach, overall management of the solution is liberated from belief and non needed separation (Silo). Then new or disruptive ideas can emerge (like using passenger device to extend the capacity of processing of an on-board system or using a personal mobile device to carry data from the gate to the aircraft)