13 November 2020

Written by Cristóvão Cordeiro, R&D Project Manager, at SixSq

With the increasing adoption of ubiquitous computing, terms like Cloud, Edge, Fog and IoT are now more popular than ever. Recent news has shown us that data is everywhere - it comes from nature, from computer analysis, from us humans… and it is being used to generate information that turns out to be critical in the most diverse sectors, from education, to healthcare, automotive and even politics.

Recent forecasts estimate that there will be approximately 41.6 billion connected IoT devices, generating 79.4 zettabytes (ZB) of data, by 2025. To put this number into perspective, one would need to binge watch 3.5 billion years of YouTube videos in order to consume this much data.

Even though data is the new gold, without a purpose, it is just white noise. Data needs processing, and this ever-increasing deluge raises new computing concerns with respect to: cost, network latency & reliability, and privacy. This is where ELASTIC's Edge-to-Cloud infrastructure comes into play.


The Edge-to-Cloud data flow model

From a management perspective, ELASTIC is putting together an intelligent analytics-driven orchestration engine which is capable of reaching out to both edge and cloud infrastructures, to deploy and distribute workloads closer to the data sources.

Take ELASTIC's smart mobility use cases as an example: all data which is critical for assisting public transport drivers is generated at the edge, thus it is only logical to process that same data at its source, without saturating the network, only keeping the actual information and results of the computation for later analysis. By doing so, ELASTIC is dramatically reducing the cost of the network, while ensuring that whatever applications run at the edge are not impacted by unexpected network delays. Plus, since Artificial Intelligence is being used for processing heavy data (like video feeds), the self-contained edge workloads ensure that the privacy bubble does not burst, and only anonymized data can travel via public networks up to the cloud.

In fact, ELASTIC's Software and Fog Architectures have been designed in such a way that they build on top of proven technologies, providing benefits for the Edge-to-Cloud and Data-to-Information paradigms. In particular, ELASTIC's Fog Computing architecture encompasses the best of two Edge Computing software appliances – KonnektBox and NuvlaBox – to deliver the best possible conditions for a seamless and performant transformation of data into useful information:

- Container-based Software Stack: these edge computing appliances are built with containers, for containers. In ubiquitous computing, monolithic and tailored applications are no longer an option. Via the use of containers, ELASTIC provides support for portable, stateless and loosely-coupled applications, thus widening the applicability of the framework to endless use cases.
- Proactive Monitoring: monitoring is critical. Moreover, when the availability of the edge devices cannot be guaranteed (due to mobile environments, unreliable edge networks, etc.), telemetry and monitoring information of the edge system becomes paramount in the identification of problems, such as underperformance and intrusion. Both the KonnektBox and NuvlaBox are equipped with edge monitoring agents capable of providing the right metrics for assessing the state of both the edge devices and the underlying applications.
- Distributed Storage: only keep what's necessary. It would be impossible to persist all the raw data coming from IoT sensors. Thus, the ELASTIC Fog Architecture comprises the mechanisms to not only store but also to flag and synchronize important data throughout the edge and up to the cloud.
- Data Routing: if it is the information that matters, then why store everything? Instead, let the application consume the intended raw data in flight, and discard everything else. The KonnektBox and NuvlaBox provide a functionality to allow the ELASTIC use cases to build applications without the need to individually and natively interact with every data source (sensor). The data router abstracts this interaction, by automatically discovering IoT peripherals and relaying their raw data through an MQTT broker, to which any authorized ELASTIC application can subscribe.