Skip to content

Conversation

@fikunmi-ap
Copy link

Problem

The current banking-bench test suite is not representative (enough) of what mainnet workloads are. Because of this, it's difficult to evaluate the scheduler before it hits a production network and even evaluating modifications is tricky.
This necessitates a more comprehensive test suite.

Summary of Changes

In addtion to the WriteLockContention enum that currently exists, I've implemented an account-pool contention algorithm that allows for more granular contention and more representative workloads by varying the number of accounts.

A quick run showed some interesting behavior that the current test suite does not of the scheduler and the greedy scheduler:

Agave Perf
Agave Perf2

@mergify mergify bot requested a review from a team March 24, 2025 21:58
@appetrosyan appetrosyan force-pushed the accounts-banking-bench branch from 5b74ceb to 915f79f Compare March 24, 2025 22:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant