Skip to content

Ingestion lock optimization using caching strategy#852

Open
VoletiRam wants to merge 8 commits intovalkey-io:mainfrom
VoletiRam:caching_radix
Open

Ingestion lock optimization using caching strategy#852
VoletiRam wants to merge 8 commits intovalkey-io:mainfrom
VoletiRam:caching_radix

Conversation

@VoletiRam
Copy link
Collaborator

Ingestion lock optimization using caching strategy

Ram Prasad Voleti added 8 commits March 3, 2026 05:13
Replace the single coarse mutex protecting all text index writes with a two-level scheme:
a sharded per-word bucket mutex pool for Postings updates and an absl reader-writer lock
for tree structural changes. This allows concurrent writes to independent words to proceed
in parallel rather than serializing all writer threads. Search reads now hold a shared
reader lock and no longer block concurrent writes to existing words. The tree write
lock is only acquired when a word is first inserted or fully removed.

Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Remove reader mutex lock on search path. Refactor and rename

Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Add gloabl shared cache abd apply serially at the end to the tree via
callback

Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Fix clang issue

Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Fix clang 2

Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Signed-off-by: Ram Prasad Voleti <ramvolet@amazon.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant