Skip to content

[runtime/iobuf/pool] improve buffer pool defaults #3860

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
patrick-ogrady merged 10 commits into from
May 22, 2026
+66 −89
Merged

[runtime/iobuf/pool] improve buffer pool defaults #3860

Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
ffff1c5
[runtime/iobuf/pool] improve buffer pool defaults
andresilva May 22, 2026
cb866e7
[runtime/iobuf/pool] remove unused
andresilva May 22, 2026
a2d8b1c
[runtime/iobuf/pool] docs
andresilva May 22, 2026
0921924
[runtime/iobuf/pool] fix tests
andresilva May 22, 2026
f865e52
[runtime/iobuf/pool] add todo
andresilva May 22, 2026
afac103
[runtime/iobuf/pool] fix test
andresilva May 22, 2026
970afb3
nit
patrick-ogrady May 22, 2026
4a32127
fmt
patrick-ogrady May 22, 2026
82dac4d
move
patrick-ogrady May 22, 2026
05605a6
cleanup
patrick-ogrady May 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 5 additions & 6 deletions runtime/src/iobuf/pool.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -192,14 +192,13 @@ impl BufferPoolConfig {
/// Cache-line alignment is used because network buffers don't require page
/// alignment for DMA, and smaller alignment reduces internal fragmentation.
pub const fn for_network() -> Self {
let cache_line = NZUsize!(cache_line_size());
Self {
pool_min_size: 1024,
pool_min_size: 0,
Copy link
Copy Markdown
Member Author

@andresilva andresilva May 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The pool is now faster than going to the system allocator, so there's no reason to bypass it in the defaults.

Comment thread
patrick-ogrady marked this conversation as resolved.
min_size: NZUsize!(1024),
max_size: NZUsize!(64 * 1024),
max_size: NZUsize!(128 * 1024),
Copy link
Copy Markdown
Member Author

@andresilva andresilva May 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not? This means this pool will use 1GB of memory at peak, same as the storage pool below.

max_per_class: NZU32!(4096),
prefill: false,
alignment: cache_line,
alignment: NZUsize!(1),
Copy link
Copy Markdown
Member Author

@andresilva andresilva May 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We don't actually need any specific alignment here, and while this doesn't make any difference once the buffer is allocated and pooled the initial allocation is slower.

parallelism: NZUsize!(1),
thread_cache_config: BufferPoolThreadCacheConfig::Enabled(None),
}
Expand All @@ -212,12 +211,12 @@ impl BufferPoolConfig {
pub fn for_storage() -> Self {
let page = NZUsize!(page_size());
Self {
pool_min_size: 1024,
pool_min_size: 0,
Copy link
Copy Markdown
Member Author

@andresilva andresilva May 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as above.

min_size: page,
max_size: NZUsize!(8 * 1024 * 1024),
max_per_class: NZU32!(64),
prefill: false,
alignment: page,
alignment: NZUsize!(1),
Comment thread
patrick-ogrady marked this conversation as resolved.
Comment thread
cursor[bot] marked this conversation as resolved.
parallelism: NZUsize!(1),
thread_cache_config: BufferPoolThreadCacheConfig::Enabled(None),
}
Expand Down
Loading