NCU

33 views
Skip to first unread message

Giant Hand

unread,
Feb 12, 2025, 9:55:34 PMFeb 12
to Google cluster data - discussions

I understand that the average_usage.cpus value in the cluster data represents NCU.
Since NCU is a normalized compute unit, shouldn't its values be confined to the 0–1 range when expressed as a percentage?
However, I noticed that some values exceed 1 (for example 65.5), and I would like to understand why this occurs.

Additionally, I would like to use the average_usage.cpus values as non-normalized (absolute) compute usage values rather than normalized percentages.
Is this possible? If so, how can I retrieve or calculate the unnormalized values?

I would appreciate your clarification on these points.

Cheers, 

john wilkes

unread,
Feb 14, 2025, 12:02:41 AMFeb 14
to googlecluste...@googlegroups.com
Hi.

From the documentation: "Normalized Compute Units, or NCUs ... values will be in the range [0,1]."  There's always a bit of noise in large-scale traces like this, but I would be surprised if there were a lot of values > 1.0  (these are fractions, not percentages!)  Do you see a large number of such values?

We do not provide information on how to denormalize the values.  (See the section on Obfuscation techniques in the documentation.)

Best,
  john


--
You received this message because you are subscribed to the "Google cluster data - discussions" group. To post to this group, send email to googlecluste...@googlegroups.com. To unsubscribe from this group, send email to googleclusterdata-...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/googleclusterdata-discuss?hl=en-US.
---
You received this message because you are subscribed to the Google Groups "Google cluster data - discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to googleclusterdata-...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/googleclusterdata-discuss/12a8b7c5-13f6-4731-8f53-940bb8b0b0e1n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages