[Tarantool-patches] [PATCH 3/4] memtx: implement api for memory allocator selection

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Sun Jan 10 16:56:25 MSK 2021

Thanks for the patch!

On 29.12.2020 12:03, mechanik20051988 via Tarantool-patches wrote:
> From: mechanik20051988 <mechanik20.05.1988 at gmail.com>
> Slab allocator, which is used for tuples allocation,
> has a certain disadvantage - it tends to unresolvable
> fragmentation on certain workloads (size migration).
> New option allows to select the appropriate
> allocator if necessary.
> @TarantoolBot document
> Title: Add new 'allocator' option
> Add new 'allocator' option which allows to
> select the appropriate allocator for memtx
> tuples if necessary.

- Option for what? box.cfg?

- What are the option values?

- How a user is supposed to choose one? Depending on what?

- Is system allocator restricted by the same memory quota?

- Does system allocator allocate all the memory at start, like
small does?

- The option introduction could be a separate commit. Now you
did refactoring + new 'feature' in one commit.

> Closes #5419

- It does not really 'close' the issue, because at this commit the
new allocator type is not implemented.

- Besides, this line is below docbot request, which means it
will go to the doc ticket. It shouldn't.

Taking into account some comments being about system allocator, it
seems you should better extract the box.cfg change into a new
commit, move it to the end, and add the docbot request to there.

More information about the Tarantool-patches mailing list