Skip to content
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

Jump to bottom

Opt #110 Optimize memory cache's singleflight performance #111

Closed
wants to merge 4 commits into from
Closed

Opt #110 Optimize memory cache's singleflight performance #111

Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f977b3f
Optimize memory cache's singleflight performance
Mar 5, 2024
755e5bc
Optimize memory cache's singleflight performance
Mar 5, 2024
d9ff3e8
just double check before lookup action
Mar 6, 2024
95ec0b5
conn pool
Jun 2, 2024
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
9 changes: 9 additions & 0 deletions pingora-memory-cache/src/read_through.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,6 +181,15 @@ where
}
None => {
let mut lockers = self.lockers.write();
// Acquiring write locks is not that easy in a high-concurrency environment,
// perhaps another one has already completed the lookup and removed the lock from locker.
// So we first check whether data already exists at the current point
let (result, cache_state) = self.inner.get(key);
if let Some(result) = result {
/* cache hit */
return (Ok(result), cache_state);
}

Copy link

Choose a reason for hiding this comment

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

If I understand correct, this is the case following match lockers.get(&hashed_key) code logic can handle, if another one already added a same hashed_key just before lockers.write().

Copy link
Author

Choose a reason for hiding this comment

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

suppose there are two threads(coroutine) t1 t2
t1 get data -> None
t1 get container read lock, get CacheLock -> None
t1 get container write lock, insert CacheLock
t1 lookup data
t2 get data -> None
t2 pause for some reason such as thread switching
t1 put lookuped data
t1 get container write lock, remove CacheLock
t2 resume and get container read lock, get CacheLock -> None
t2 get container write lock, insert another CacheLock
t2 lookup data

So it is waste that t2 lookup data in this situation.
Double check should be done when get write lock like thread-safe lazy initialization.

match lockers.get(&hashed_key) {
Some(lock) => {
/* another lookup to the same key got the write lock to locker first */
Expand Down