Are you asking how Airtable stores its underlying data? Airtable has not published that information, and I doubt that it is any of those you listed. And even if they did, you cannot use any of those tools to work directly with Airtable data.
You cannot use SQL directly with an Airtable base. However, Airtable is a relational database. It just has its own unique way of dealing with relationships—via linked record fields. Even though you cannot use SQL, I wouldn’t call it a NoSQL platform.
The structure of a table in Airtable is what you define it to be, including its fields, views, and permission settings. And view settings covers a huge amount of features.
>>> It is better to abstract the data and put it into few tables (Associated by some fields)?
No. It is better to just use Airtable without the need to consider the underlying storage architecture. Airtable is designed for people who have no time or skills to ponder such questions. If you believe you are going to somehow alter or enhance Airtable's behaviour based on such deeper knowledge, you are probably going to be disappointed.
>>> So there will be tens of millions of data tables in database?
Even if that were the case, why would you care?
At the outset, it's very unlikely that Airtable comingles data from customer (a) with customer (b) because meeting certain security compliance certifications would rule that out. I'm almost certain your data is not stored in a common tenancy architecture. Furthermore, since there is good evidence that the various service levels exist, there's likely a container architecture running with your instance of Airtable and data storage.
Even though the thread is from a year ago, the topic of databases remains relevant and exciting! In Airtable, the database used is NoSQL, specifically MongoDB. This choice provides a flexible and dynamic structure for organizing data efficiently.
Airtable's tables consist of various fields, allowing you to customize data types and capture different information like text, numbers, dates, attachments, and more. Views in Airtable enable you to organize and filter data in different ways, enhancing your productivity and decision-making.
For those interested in exploring the art of designing a database, check out this database design service. This resource can provide valuable insights and principles to help you optimize your database structure and improve overall performance.
@carrgordon wrote: the database used is NoSQL, specifically MongoDB. This choice provides a flexible and dynamic structure for organizing data efficiently.
But comes with great limitations. Airtable lacks any ability to build and manage documents. A lot of money, development, and added moving parts burden developers who need their data to blend well with documents. Their choice of NoSQL rules out certain architectural agility such as rich document management and this will ultimately cause other problems.
A good example is AI. Not many of us realize (or care) about managing vectors. But this has emerged as the number one design choice for anyone who wants to use semantic similarities for finding, sorting, and discovering patterns using generative AI. Have you ever asked why Airtable doesn't have a pervasive search feature that can find stuff without using exact terms and data?
Text embedding vectors can be added to NoSQL databases, but many issues exist. SaaS platforms based on PostgreSQL are wading into the AI pool effortlessly because it is well-positioned to serve as a vector database.
NoSQL means Not Only SQL. Airtable has no reporting language because, to them, they never surfaced SQL as part of the offering. And they perceived a market where SQL was actually unimportant. With their new enterprise focus, I have a hunch many customers have asked about SQL, so there's that.