
We have been seeing it in movies for years, cars that drive without any human intervention or interaction. What looks futuristic at first, is now being worked on by all car manufacturers. The ultimate goal is building a full Level 5 Autonomous Car. Though it might still take a couple of years (or decades) to achieve this, the technology investments are already happening today and the foundations are laid for these challenging projects.
In order to build such cars and get these validated, massive data sets need to be processed and analyzed over an over again. The industry has set clear expectations and the validation of such “human replacement” scenarios are lengthy and very expensive. The goal is clear, build a car that can do better than a human … but can it?
We will start by talking about the business side of things. What are the 5 levels of autonomous cars, where are we today and what are the (legal) requirements to bring these to market. We will also clearly sketch the challenges the industry is faced with and what they need to come up with in order to succeed over time.
In the major part of the talk, we will go deeper into the architectural side of things. We will walk you through the complete end-to-end “reference” architecture of such solutions on Azure. Don’t expect too many demos or code, but a true architectural talk on the challenges and solutions for massive scale. First we will explore the mass data ingestion challenge, expected to be in the PB range on a day to day basis. We will talk about the challenges around data redaction and master data management, in order to be allowed to use the data. Once these 2 pieces have been cleared out, we will go deeper on the mass compute and storage needs and the appropriate Azure architecture that can be leveraged to successfully build these kinds of solutions.
To conclude, we will see a couple of these Azure services in a concrete and life scenario, like VMSS, Limitless Storage / ADLS v2, Azure Cycle and many more.