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

Sync fork #8

Open
wants to merge 115 commits into
base: master
Choose a base branch
from
Open

Sync fork #8

wants to merge 115 commits into from

Conversation

smihaylov-crystal
Copy link

No description provided.

mdashti and others added 30 commits February 21, 2024 00:45
@mdashti @smihaylov-crystal
Added initial_size + Added user/pass to connect + NItz
9f9f7ce 
Co-authored-by: smihaylov-crystal <[email protected]>
@mdashti @smihaylov-crystal
Added autovac.
f811314 
Co-authored-by: smihaylov-crystal <[email protected]>
@smihaylov-crystal
pid_demo
bf9e806
@smihaylov-crystal
rl initial commit
9127153
@smihaylov-crystal
numerious fixes, incl transition to discrete control of autovacuum
acf6e7a
@smihaylov-crystal
Wait until initial data is populated
3780894
@smihaylov-crystal
tweaks
7cd0c5d
@smihaylov-crystal
support passing user, password, url and dbname as args
6c29005
@smihaylov-crystal
fixes to set benchmarking options correctly, and tweaks to reward fun…
ff5c7a3 
…ction
@smihaylov-crystal
specify number of episodes from command line
86e8e70
@smihaylov-crystal
Add support for testing a previously learned model
0c30ad4
@smihaylov-crystal
Make autovacuum penalty smaller
f10ab60
@smihaylov-crystal
Update reward. Do not penalize for a lot of data being read
172db18
@smihaylov-crystal
Ignore updating stats when the table hasnt been created yet.
d91384b
@smihaylov-crystal
Refactor RL code to better support custom workloads
d854264
@smihaylov-crystal
tweaks
9d49d7b
@smihaylov-crystal
tweaks and formatting
c51b5b7
@smihaylov-crystal
restore statement and small refactor for testability
d79b92f
@smihaylov-crystal
fix import
da9d792
@smihaylov-crystal
Make sure each action has at least a small chance of being selected w…
2f3138e 
…hen learning.
@smihaylov-crystal
1. Update to reward function.
7cfa77a 
2. Normalize state, keep domain small
3. Randomize experiments (run with different parameters)
@smihaylov-crystal
Use pow() instead of exp2()
65af791
@smihaylov-crystal
Use smaller transaction size
e83a06d
@smihaylov-crystal
comment out action randomizing code
f1501d1
@smihaylov-crystal
Refactor PG-specific logic outside of RL
41276ea
@smihaylov-crystal
Make sure autovacuum is disabled
a5d5df9
@smihaylov-crystal
Tweaks
4e6562b
@smihaylov-crystal
Initial version of simulated autovacuum
a601c1d
@smihaylov-crystal
tweaks and formatting
71d587d
@smihaylov-crystal
autovac demo
9543084
@smihaylov-crystal
Record action in replay files.
c21be01
@smihaylov-crystal
Separate connection for vacuuming to support async
95a8dc1
@smihaylov-crystal
Disable separate vacuum thread for now
c2d8222
@smihaylov-crystal
Add connection lock
cd85bc0
@smihaylov-crystal
Dedicated connection for vacuuming
50d18fc
@smihaylov-crystal
Disable
3f432dd
@smihaylov-crystal
Vacuuming in separate thread.
3685fb2
@smihaylov-crystal
Restore checkpoint format. Needs more than the state dict.
25493b1
@smihaylov-crystal
Tweak rewards
509ec73
@smihaylov-crystal
Flush stdout
e39e271
@smihaylov-crystal
Record vacuuming history in state instead of query perf
d001ebf
@smihaylov-crystal
Tweak replay buffer sizes
34e7a4c
@smihaylov-crystal
* add ability to resume a model (finetune)
59786f7 
* decrease learning rate for stability
* randomize initial state
@smihaylov-crystal
Set correct experiment_id when finetuning a model
8a0f620
@smihaylov-crystal
Increment experiment_id via environment instead of via executors
f906063
@smihaylov-crystal
* Make performance-based reward relative to 90th percentile
8a9cc09 
* Tweaks to model save files
@smihaylov-crystal
Avoid NAN. Tweak learning params.
426731a
@smihaylov-crystal
Small refactor
db7bc06
@smihaylov-crystal
Refactor move finetuning logic inside agent
5cbd9a6
@smihaylov-crystal
Make training optional
f094a09
@smihaylov-crystal
Small refactor
ac930b3
@smihaylov-crystal
More appropriate class names
3d1f120
@smihaylov-crystal
Add PID agent
462f625
@smihaylov-crystal
Small fix for simulated free space
6d5618d
@smihaylov-crystal
Fix asserts
00ead73
@smihaylov-crystal
Implement no-op agent with selectable default action.
63a045c
@smihaylov-crystal
Decouple learning and benchmarking code from iibench.py
f837331
@smihaylov-crystal
More work to decouple workload from pg_stat_and_vacuum.
a022c08
@smihaylov-crystal
Remove barrier from ibench code. Instead, use a query to establish wo…
d55be79 
…rkload is initialized.
@smihaylov-crystal
Remove unused script
f636009
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@smihaylov-crystal @mdashti @jssmith