Skip to content
This repository has been archived by the owner on Aug 2, 2023. It is now read-only.

More detailed resource usage statistics #103

Closed
4 tasks done
achimnol opened this issue Oct 24, 2018 · 0 comments
Closed
4 tasks done

More detailed resource usage statistics #103

achimnol opened this issue Oct 24, 2018 · 0 comments
Assignees
Milestone

Comments

@achimnol
Copy link
Member

achimnol commented Oct 24, 2018

Let's provide three-tuple of statistics for each resource slot type, used / allocated / total, in the admin GraphQL API.

e.g., If an agent has 8 GiB memory and a kernel session is allocated 4 GiB but using only 2.5 GB, we could show a resource gauge like this in a GUI-based console:

NOTE: Later, a single value of absolute "resource slot" in the current code will become a pair of "amount" (absolute real values such as cores and bytes) and "share" (relative virtual decimal values).

  • Let agents report detailed stats upon heartbeats.
  • Extend the GraphQL API to provide all three used/alloacted/total values
  • Deprecate "overbooking" factors -- we are now going to optimize resource limits using real usage traces and via policies
  • Update the API documentation
@achimnol achimnol added this to the 1.5 milestone Oct 24, 2018
@achimnol achimnol changed the title Usage statistics More detailed resource usage statistics Oct 24, 2018
@adrysn adrysn assigned adrysn and unassigned inureyes Dec 6, 2018
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 8, 2018
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 9, 2018
* More easy to understand naming for CPU and memory display names.
adrysn added a commit that referenced this issue Dec 9, 2018
* Calculating CPU % is not needed anymore, since it is deletaged to agent during heartbeat. It is stored in redis stats server.
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 10, 2018
adrysn added a commit that referenced this issue Dec 10, 2018
* It is not useful to calculate cpu % for a specific container. Displaying cpu used (ns) will be sufficient.
adrysn added a commit that referenced this issue Dec 11, 2018
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 11, 2018
adrysn added a commit that referenced this issue Dec 13, 2018
* Current cpu and memory usages (cpu_cur_pct and mem_cur_bytes) are delivered when resolving agent query.
* Remove overbook factors.
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 13, 2018
adrysn added a commit to lablup/backend.ai-client-py that referenced this issue Dec 13, 2018
@achimnol achimnol modified the milestones: 18.12, 19.03 Apr 11, 2019
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

No branches or pull requests

3 participants