aws-performance

name	aws-performance
description	Monitor and analyze AWS infrastructure performance using read-only MCP tools
version	1.0.0
author	kai-agent
metadata	{"kai":{"tags":["kai","aws","performance","monitoring","cloudwatch"],"related_skills":["aws-cost-optimizer"]}}

AWS Performance Analyzer

Monitor health and performance of AWS infrastructure. All operations are read-only — no changes are made.

Prerequisites

• Workspace with AWS credentials configured in kai-backend
• MCP tools available: mcp__kai__aws_*

Workflow 1: Infrastructure Overview

Get a complete picture of what's running.

aws_ec2_list_instances(workspaceId, state="running")     # Compute
aws_ecs_list_services(workspaceId)                        # Containers
aws_lambda_list_functions(workspaceId)                    # Serverless
aws_rds_list_instances(workspaceId)                       # Databases
aws_elb_list(workspaceId)                                 # Load balancers
aws_cloudformation_list_stacks(workspaceId)               # IaC stacks

Present as a summary table:

Service	Count	Key Details

Workflow 2: ECS Health Check

1. Service status:

   aws_ecs_list_services(workspaceId)
   

   • Flag services where runningCount < desiredCount
   • Flag services with pendingCount > 0 for extended time

2. Recently crashed tasks:

   aws_ecs_list_tasks(workspaceId, desiredStatus="STOPPED")
   

   • Check stoppedReason for OOM kills or task failures

3. Resource utilization (CPU + Memory):

   aws_cloudwatch_get_metrics(
     workspaceId, namespace="AWS/ECS", metricName="CPUUtilization",
     dimensions=[{"name":"ClusterName","value":"xxx"},{"name":"ServiceName","value":"yyy"}],
     startTime="-24h", endTime="now", period=300
   )
   

   Repeat for MemoryUtilization.

   • Flag CPU sustained >80% → upgrade candidates
   • Flag Memory sustained >90% → OOM risk

Workflow 3: EC2 Performance Analysis

For each running instance:

aws_cloudwatch_get_metrics(workspaceId, namespace="AWS/EC2", metricName="CPUUtilization",
  dimensions=[{"name":"InstanceId","value":"i-xxx"}],
  startTime="-7d", endTime="now", period=3600, statistics=["Average","Maximum"])

Also check:

• NetworkIn / NetworkOut — network bottlenecks
• EBSReadOps / EBSWriteOps — disk I/O pressure

Decision matrix:

Avg CPU	Max CPU	Recommendation
<10%	<30%	Downsize or stop
10-60%	<80%	Right-sized
>60%	>80%	Consider upgrading
>80%	>95%	Upgrade urgently

Workflow 4: Lambda Performance

1. List functions:

   aws_lambda_list_functions(workspaceId)
   

2. For each function, check metrics:

   aws_cloudwatch_get_metrics(workspaceId, namespace="AWS/Lambda",
     metricName="Duration",  # also: Errors, Throttles, ConcurrentExecutions
     dimensions=[{"name":"FunctionName","value":"xxx"}],
     startTime="-7d", endTime="now", period=86400, statistics=["Average","Maximum","Sum"])
   

   Flags:

   • Error rate >5% → investigate immediately
   • P99 duration >80% of timeout → at risk of timing out
   • Throttles >0 → hitting concurrency limits
   • ConcurrentExecutions near account limit → scale risk

Workflow 5: Database Health (RDS)

1. List databases:

   aws_rds_list_instances(workspaceId)
   

2. For each DB, check CloudWatch:

   • CPUUtilization — sustained >80% is a problem
   • DatabaseConnections — compare to max_connections setting
   • FreeableMemory — low = swap risk
   • ReadIOPS / WriteIOPS — I/O bottlenecks

   aws_cloudwatch_get_metrics(workspaceId, namespace="AWS/RDS",
     metricName="CPUUtilization",
     dimensions=[{"name":"DBInstanceIdentifier","value":"xxx"}],
     startTime="-24h", endTime="now", period=300)
   

Workflow 6: Load Balancer Health

1. List LBs:

   aws_elb_list(workspaceId)
   

2. Check target health:

   aws_elb_target_health(workspaceId, targetGroupArn="arn:xxx")
   

   • Flag any unhealthy targets
   • Flag targets in "draining" state

3. Check latency (CloudWatch):

   aws_cloudwatch_get_metrics(workspaceId, namespace="AWS/ApplicationELB",
     metricName="TargetResponseTime",
     dimensions=[{"name":"LoadBalancer","value":"app/xxx/yyy"}],
     startTime="-24h", endTime="now", period=300, statistics=["Average","Maximum"])
   

   • Flag avg response time >1s
   • Check HTTPCode_Target_5XX_Count for error spikes

Workflow 7: Security Posture Quick Check

1. Security groups:

   aws_ec2_security_groups(workspaceId)
   

   • Flag rules with 0.0.0.0/0 on ports other than 80, 443
   • Flag rules with 0.0.0.0/0 on SSH (22) or RDP (3389)

2. IAM users:

   aws_iam_list_users(workspaceId)
   

   • Flag users with passwordLastUsed: "never" — unused accounts

3. IAM roles overview:

   aws_iam_list_roles(workspaceId)
   

CloudWatch Metrics Quick Reference

Service	Namespace	Common Metrics	Dimension
EC2	AWS/EC2	CPUUtilization, NetworkIn, NetworkOut, DiskReadOps	InstanceId
ECS	AWS/ECS	CPUUtilization, MemoryUtilization	ClusterName, ServiceName
Lambda	AWS/Lambda	Duration, Errors, Throttles, Invocations, ConcurrentExecutions	FunctionName
RDS	AWS/RDS	CPUUtilization, DatabaseConnections, FreeableMemory, ReadIOPS, WriteIOPS	DBInstanceIdentifier
ALB	AWS/ApplicationELB	TargetResponseTime, HTTPCode_Target_5XX_Count, RequestCount	LoadBalancer
NLB	AWS/NetworkELB	ActiveFlowCount, ProcessedBytes	LoadBalancer
S3	AWS/S3	BucketSizeBytes, NumberOfObjects	BucketName, StorageType

Reporting Format

Infrastructure Health Report

Overall status: Healthy / Warning / Critical

Resource	Service	Status	Issue	Severity
i-abc123	EC2	Warning	CPU avg 85% over 7d	High
my-service	ECS	Critical	2/4 tasks running	Critical
auth-fn	Lambda	Warning	Error rate 8%	High

Immediate actions needed:

1. ...

Monitoring recommendations:

1. ...