This repository has been archived by the owner on Jan 23, 2023. It is now read-only.
Partially improve support for --cpus from Docker CLI
#23747
Merged
Conversation
This file contains 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
jkotas
approved these changes
Apr 4, 2019
janvorli
approved these changes
Apr 4, 2019
kouvel
approved these changes
Apr 4, 2019
tmds
approved these changes
Apr 5, 2019
tmds
reviewed
Apr 5, 2019
In the case where `--cpus` is set to a value very close to the smaller integer (ex: 1.499999999), it would previously be rounded down. This would mean that the runtime would only try to take advantage of 1 CPU in this example, leading to underutilization. By rounding it up, we augment the pressure on the OS threads scheduler, but even in the worst case scenario (`--cpus=1.000000001` previously being rounded to 1, now rounded to 2), we do not observe any overutilization of the CPU leading to performance degradation.
By making sure we do take the CPU limits into account when computing the CPU busy time, we ensure we do not have the various heuristic of the threadpool competing with each other: one trying to allocate more threads to increase the CPU busy time, and the other one trying to allocate less threads because there adding more doesn't improve the throughput. Let's take the example of a system with 20 cores, and a docker container with `--cpus=2`. It would mean the total CPU usage of the machine is 2000%, while the CPU limit is 200%. Because the OS scheduler would never allocate more than 200% of its total CPU budget to the docker container, the CPU busy time would never get over 200%. From `PAL_GetCpuBusyTime`, this would indicate that we threadpool threads are mostly doing non-CPU bound work, meaning we could launch more threads.
sergiy-k
reviewed
Apr 5, 2019
| cpu_count = quota / period; | ||
| if (cpu_count < UINT32_MAX) | ||
| cpu_count = (double) quota / period; | ||
| if (cpu_count < UINT32_MAX - 1) |
There was a problem hiding this comment.
Should this be '<=' otherwise if cpu_count is equal to (UINT32_MAX - 1) then this function will return UINT32_MAX instead of (UINT32_MAX - 1), right?
In general, I agree with @tmds here that rounding up before the 'if' statement might make the code simple. Would something like this work or I missed some details?
// Calculate cpu count based on quota and round it up
cpu_count = (double) quota / period + 0.999999999;
*val = (cpu_count < UINT32_MAX) ? (uint32_t)cpu_count : UINT32_MAX;
There was a problem hiding this comment.
It's ok to return UINT32_MAX. I didn't adopt @tmds approach mostly to reduce code change noise.
|
In which dotnet core runtime version is this PR available? |
|
It is available in the .NET Core 3.0 preview6. https://dotnet.microsoft.com/download/dotnet-core/3.0 |
picenka21
pushed a commit
to picenka21/runtime
that referenced
this pull request
Feb 18, 2022
…r#23747) * Round up the value of the CPU limit In the case where `--cpus` is set to a value very close to the smaller integer (ex: 1.499999999), it would previously be rounded down. This would mean that the runtime would only try to take advantage of 1 CPU in this example, leading to underutilization. By rounding it up, we augment the pressure on the OS threads scheduler, but even in the worst case scenario (`--cpus=1.000000001` previously being rounded to 1, now rounded to 2), we do not observe any overutilization of the CPU leading to performance degradation. * Teach the ThreadPool of CPU limits By making sure we do take the CPU limits into account when computing the CPU busy time, we ensure we do not have the various heuristic of the threadpool competing with each other: one trying to allocate more threads to increase the CPU busy time, and the other one trying to allocate less threads because there adding more doesn't improve the throughput. Let's take the example of a system with 20 cores, and a docker container with `--cpus=2`. It would mean the total CPU usage of the machine is 2000%, while the CPU limit is 200%. Because the OS scheduler would never allocate more than 200% of its total CPU budget to the docker container, the CPU busy time would never get over 200%. From `PAL_GetCpuBusyTime`, this would indicate that we threadpool threads are mostly doing non-CPU bound work, meaning we could launch more threads. Commit migrated from dotnet/coreclr@aea3b1a
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
This focuses on better supporting Docker CLI's parameter
--cpus, which limits the amount of CPU time available to the container (ex: 1.8 means 180% CPU time, ie on 2 cores 90% for each core, on 4 cores 45% on each core, etc.) in the case of the ThreadPool and in the case of calculating the CPU limit.