You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Feb 26, 2021. It is now read-only.
Improvements to Npgsql:
Connection pool contention optimization (@roji to file issues on possible improvements for ASP.NET Core pooling code)
Removal of Task allocations
Prepared statements written directly to buffer (still pending)
Possible improvements for ADO.NET
API Connection pooling could be made explicit (expose a connection pool) to avoid the lookup and the allocation of the user facing connection
Explicit prepared commands that allows reusing commands definitions
Prepare (as well as other APIs) needs to be async
Reader pooling
Update
Ability to write parameters without boxing (Npgsql has a non ADO.NET API for it)
Other update performance such as batching
Possible improvements for EF Core:
Consider batching for query
Long term goals
Apply same optimizations to other ADO.NET providers
We can experiment with pipelines, but later would be better (could help with SSL)
The text was updated successfully, but these errors were encountered:
Attendees: @roji @ajcvickers @anpete @sebastienros @divega @davidfowl @DamianEdwards @mikeharder
Npgsql improvements on Fortunes for physical hardware on Linux:
60,000 rps - Npgsql Latest released
77,000 rps - Npgsql New release
97,000 rps - Npgsql latest experimental changes
117,000 rps - Peregrine
119,000 rps - Undertow (Java)
TO-DO:
Improvements to Npgsql:
Connection pool contention optimization (@roji to file issues on possible improvements for ASP.NET Core pooling code)
Removal of Task allocations
Prepared statements written directly to buffer (still pending)
Possible improvements for ADO.NET
Possible improvements for EF Core:
Long term goals
The text was updated successfully, but these errors were encountered: