AgentaaS OS
IFO4 PLAYGROUND
P1
1
Pull Namespace Cost Breakdown2
Identify Top Cost NamespacesP2
3
Diagnose Pod Resource Usage4
Find Missing Resource Limits5
Identify Node Type IssuesP3
6
Set Resource Requests and Limits7
Implement LimitRanges8
Configure HPAP4
9
Configure Kubecost10
Set Namespace BudgetsP5
11
Move Dev to Spot Nodes12
Configure Cluster AutoscalerP6
13
Set AgentaaS Alerts14
Create Team DashboardInitiatives5
Capital Under Change$7.8M
Health Score58/100
Waste %34.2%
Value at Risk$6.7M
Phase 1: Discovery
Pull Namespace Cost Breakdown
ANALYTICS15 pts
Spend Trend
K8s Namespace Costs
ml-experiments
47 pods · CPU 312/847 vCPU
$53K
no limitsproduction-api
84 pods · CPU 398/420 vCPU
$18K
data-pipeline
38 pods · CPU 162/180 vCPU
$13K
monitoring
22 pods · CPU 34/40 vCPU
$5K
SITUATION
Pull the namespace cost breakdown from the Kubecost integration in AgentaaS OS. The cluster has 4 namespaces. The ML team namespace is consuming $53,400/month - 60% of total cluster cost - but is only running 12 production models. The remaining $35,600 covers 3 production namespaces with 200+ microservices.
Health
58/100
Waste
34.2%
Spend
$89K/mo
Savings
$18K
AGENT INSIGHT
Cost Optimizer: ml-experiments is requesting 847 vCPU but using 312 (37% efficiency). With proper limits at actual usage, namespace cost drops to ~$19,800/month - a $33,600/month savings.
DECISION REQUIRED
Kubecost shows ml-experiments at $53,400/month with 847 CPU requested but only 312 vCPU actually used. What does this reveal?
Hint: Kubecost allocates cost based on resource REQUESTS. Pods without limits can request unlimited CPU.