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

Make Scan tasks use a spread scheduling strategy #1940

Closed
jaychia opened this issue Feb 21, 2024 · 0 comments · Fixed by #1950
Closed

Make Scan tasks use a spread scheduling strategy #1940

jaychia opened this issue Feb 21, 2024 · 0 comments · Fixed by #1950
Assignees
@clarkzinzow
Labels
p0 Priority 0 - to be addressed immediately

Comments

@jaychia
Copy link
Contributor

jaychia commented Feb 21, 2024 •
edited
Loading

This should be a pretty simple fix:

When scheduling any tasks involving a read, we should spread them across the cluster (similar to what we do for reduces)
@jaychia jaychia added the p0 Priority 0 - to be addressed immediately label Feb 21, 2024
@clarkzinzow clarkzinzow mentioned this issue Feb 26, 2024
Merged
clarkzinzow added a commit that referenced this issue Feb 26, 2024
@clarkzinzow
[PERF] Spread scan tasks over Ray cluster. (#1950)
1a94752 
This PR forces a `SPREAD` scheduling strategy for scan tasks when using
the Ray runner. This should result in better load balancing of read
tasks across the Ray cluster, yielding:
- better utilization of the aggregate network bandwidth of the cluster,
- better memory stability due to a more even post-read object
distribution,
- better performance of downstream parallel compute operations due to a
more even distribution of data over the compute bandwidth of the
cluster.

Closes #1940
samster25 pushed a commit that referenced this issue Feb 27, 2024
@clarkzinzow @samster25
[PERF] Spread scan tasks over Ray cluster. (#1950)
24a17dc 
This PR forces a `SPREAD` scheduling strategy for scan tasks when using
the Ray runner. This should result in better load balancing of read
tasks across the Ray cluster, yielding:
- better utilization of the aggregate network bandwidth of the cluster,
- better memory stability due to a more even post-read object
distribution,
- better performance of downstream parallel compute operations due to a
more even distribution of data over the compute bandwidth of the
cluster.

Closes #1940
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@clarkzinzow clarkzinzow

Labels
p0 Priority 0 - to be addressed immediately
Projects
No open projects
Daft-OSS
Archived in project
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

[PERF] Spread scan tasks over Ray cluster. Eventual-Inc/Daft
2 participants
@clarkzinzow @jaychia