Cloud, Data and AI: the three absolute buzzwords! The expectations of organizations on these topics are extremely strong because of the transformation and value creation they promise. The volume of data is exploding, the AI is driven by new disruptive technologies for CIOs, and the cloud proposes to manage all this complexity with agility! At the time of designing its Data platform (which will have to welcome all the initiatives of IA: those already imagined and especially those which are not even imaginable), the choice to build it in the cloud is very similar to a Cornelian choice. Between performance promises at a lower cost, cyber-security risk and regulatory headache, this article looks at Data and IA platforms in the cloud.
In recent years, the cloud has been at the heart of CIOs’ concerns. Widely spread over certain application areas, such as those of CRM (thanks in particular to the force of attraction of Salesforce ) or peripheral functions at the core of the business, Data projects were until now still predominantly on architectures called “on-premise”. “(The data is stored in the company’s servers and therefore located in the organization’s data centers).
The first brake on the cloud is technical. It is due to the volume of data managed in the data information system. Indeed, it has a large amount of data (several terabytes To, even petabytes Po) and it is traditionally operated in batch mode and not run-of-the-river (the data are processed batch, so with a significant volume at each treatment). This limit is becoming less and less valid thanks to the increased capacity of the networks and the capacity of the new databases and data management tools to be processed on a regular basis.
In reality, the main obstacle today is cultural. Indeed, the jump to make to accept that its data (that is to say, its war chest, potentially all the knowledge of the company!) Are hosted by a provider seems huge for many organizations. Beyond the legitimate (and wise) concern about the hot topic of data security (on which we will return), the discussion is more like a debate in the coffee trade than a reasoned study.
Cloud sirens are attractive for Data and AI projects
It must be admitted that the benefits of the cloud are numerous and very attractive, especially for Data projects:
• Cost: Usage Billing and Total Cost of Ownership (TCO) reduction, including reduced architecture administration costs
• Infrastructure: robustness, elasticity, scalability
• Methodological: super-fast startup of projects and agility of solutions. Scalability capacity
• Application: choice of solutions from a large proprietary catalog to the cloud operator or open (marketplace system)
In addition, it will be more and more common to use the frameworks of AI available in the cloud. Indeed, it is the algorithms of Google , Facebook , IBM or Microsoft , pre-trained with millions of user interactions and images, which are the most powerful and fastest to put implemented.
The cloud seems to be the eldorado of Data and AI projects. At the same time catalyst of innovation and support of the passage to the scale, it allowed many startups to see the day and to become unicorns, like Netflix , Blablacar or N26 (one of the troublemakers of 100% digital bank).
Many benefits for AI projects
The cloud also brings many benefits for AI projects, responding to specific constraints in this area:
• Large volume of data to manage> ability to have large and efficient infrastructures by separating storage and computing units (compute)
• Mobilization of important computing resources on a limited time (for the learning phase for example)> elasticity and user fees (we only pay the calculation units for the necessary time)
• Management of unstructured data (text, images, video, sound)> dedicated application solutions integrated into cloud services
• Use of specialized algorithms> service call to pre-trained algorithms of the cloud operator or other providers (interoperability and open services)
• Methodology based on iterative development and scaling> scalability and devops