Data in Python is stored in memory in a particular way. To gain an in-depth understanding on a high level, regarding how data is handled in concurrent programs, we first need to dive deep into the theoretical structure of Python memory allocation. In this section, we will discuss how data is allocated in a private heap, and the handling of this data via the Python memory manager—an overarching entity that ensures the integrity of the data.

The Python memory manager consists of a number of components that interact with different entities and support different functionalities. For example, one component handles the allocation of memory at a low level by interacting with the memory manager of the operating system that Python is running on; it is called the raw memory allocator.

On the higher levels, there are also a number of other memory allocators that interact with the aforementioned private heap of objects and values. These components of the Python memory manager handle object-specific allocations that execute memory operations that are specific to the given data and object types: integers have to be handled and managed by a different allocator to one that manages strings, or one for dictionaries or tuples. As storing and reading instructions varies between these data types, these different object-specific memory allocators are implemented to gain additional speed while sacrificing some processing space.

One step lower than the aforementioned raw memory allocator are the system allocators from the standard C library (assuming that the Python interpreter under consideration is CPython). Sometimes known as general-purpose allocators, these written-in-C entities are responsible for helping the raw memory allocator interact with the memory manager of the operating system.

The entire model of the Python memory manager described previously can be illustrated by the following diagram: