[DKG] Workloads are not configured with requests and limits for resources.

glaciation-heu / IceStream

T6.2 - Novel Metadata Fabric for Energy-Efficient Privacy Preserving Data Movement

1 stars 0 forks source link

[DKG] Workloads are not configured with requests and limits for resources. #191

Closed ktatarnikovhiro closed 1 week ago

ktatarnikovhiro commented 1 month ago

Workloads need to be configured with requests/limits explicitly for the containers with all resources: cpu/memory/ephemeral-storage/gpu

This is for those that has workloads ready: DELL/MEF

ainnocenti2011 commented 1 month ago

For Use Case 1 Sogei/Mef, the capacity information can be found in the deployment.yaml (https://github.com/glaciation-heu/mef-sog-uc1/blob/main/workload-core/deployment.yaml). The resources allocated to the PODs, including the replicas for both microservices, are configured. Currently, we are building a historical record of two months of processing on these resources to support the WorkLoadPlacement/AI process. The running is a continuous process in the Validation Environment.

ainnocenti2011 commented 1 month ago

For Use Case 1 Sogei/Mef The only unused parameter is the GPU one, as it's actually not needed: GPU - 0:0

parklize commented 1 month ago

UC2 workload service has been configured for four resources: mem, cpu, gpu, storage http://argocd.integration/applications/uc2-workload-service

ktatarnikovhiro commented 3 weeks ago

DELL is done, SOGEI is in progress