#5: Add improved outlier detection #6
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
cwschilly
reviewed
Mar 10, 2025
pierrepebay
force-pushed
the
5-add-improved-outlier-detection
branch
2 times, most recently
from
63c7091 to
fef20cd
Compare
pierrepebay
force-pushed
the
5-add-improved-outlier-detection
branch
from
8f84394 to
b48c308
Compare
pierrepebay
force-pushed
the
5-add-improved-outlier-detection
branch
from
7d178d2 to
f1aa013
Compare
lifflander
approved these changes
Mar 24, 2025
cwschilly
reviewed
Mar 24, 2025
Comment on lines
+66
to
+73
| self.expected_slow_ranks = set([ | ||
| 1702,1462,1902,1222,1182,1262,1862,1342,1382,1422,1742,1502,1102,1582, | ||
| 1142,1782,1662,1022,1062,1542,1622,982,1302,1822,1381,1501,1021,1341, | ||
| 1541,1301,1701,1821,1261,1621,1901,1012,1461,1861,1181,1221,981,1661, | ||
| 1061,1781,1421,1741,1101,1581,1141,902,1812,1252,1492,1532,1772,1292, | ||
| 1332,1852,302,382,182,702,1652,1212,582,1172,1452,1692,972,1572,1732, | ||
| 62,1412 | ||
| ]) |
There was a problem hiding this comment.
We may want to brainstorm a different way to test this so we don't have to hard-code the expected values (that's likely outside the scope of this PR though).
lifflander
requested changes
Apr 8, 2025
detection/detect_slow_nodes.py
Outdated
| threshold = representative_center + 3 * np.std(cluster_to_times[representative_cluster]) | ||
|
|
||
| problematic_clusters = [cluster_id for cluster_id, center in cluster_centers.items() if center > threshold] | ||
| return data,clusters,cluster_to_times,cluster_to_ranks,cluster_centers,representative_cluster,representative_center,threshold,problematic_clusters |
There was a problem hiding this comment.
I think instead of returning a huge tuple like this we should use a class since this is very error prone.
detection/detect_slow_nodes.py
Outdated
|
|
||
| representative_cluster = max(cluster_to_times.items(), key=lambda v: len(v[1]))[0] | ||
| representative_center = cluster_centers[representative_cluster] | ||
| threshold = representative_center + 3 * np.std(cluster_to_times[representative_cluster]) |
There was a problem hiding this comment.
Should be make this 3 a parameter instead of a constant used directly in the code?
| if representative_cluster_is_slowest: | ||
| if representative_center - 3 * np.std(cluster_to_times[representative_cluster]) > slowest_non_representative_center: | ||
| print() | ||
| print(f" WARNING: Clustering results found most times to be slower than others. No outliers will be detected.") |
There was a problem hiding this comment.
I'm thinking that instead of a warning here, we should stop the job from running by not giving the node list with a bunch of slow nodes.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes: #5