Why Performance Tuning Matters

You’ve got visibility, alerts, and dashboards. But when things slow down, or start acting weird, you need deeper data-driven troubleshooting—not guesswork.

Azure Monitor churns through a stream of telemetry, and this post is all about turning that raw data into actionable performance insights. We’ll focus on:

• Navigating and querying Azure Metrics
• Using Log Analytics for correlation and hunting
• Pinpointing performance bottlenecks before alerts fire

Metrics vs Logs: Quick Compare

You’ll often use both together metrics raise the flag, logs explain why.

Tools You’ll Use

• Azure Monitor Metrics Explorer
• Log Analytics (KQL)
• Workbooks
• Built-in metrics + diagnostic logs on services like:
  • App Services
  • AKS
  • SQL Database
  • Application Gateway

Log Query Examples for Performance Investigation

CPU Bottleneck on VMs

1
2
3
4

Perf
| where ObjectName == "Processor" and CounterName == "% Processor Time"
| summarize avg(CounterValue) by bin(TimeGenerated, 5m), Computer
| order by avg_CounterValue desc

Web App Latency Analysis

1
2
3
4

AppRequests
| where DurationMs > 1000
| summarize SlowRequests=count() by bin(TimeGenerated, 5m), Name
| order by SlowRequests desc

App Gateway Backend Errors

1
2
3

AzureDiagnostics
| where ResourceType == "APPLICATIONGATEWAYS" and BackendStatusCode startswith "5"
| summarize count() by TimeGenerated, BackendPoolName

Pair these queries with dashboards or surface them in Workbooks for fast visibility during high load.

Metrics Deep Dive: Metrics Explorer

1. Go to a resource (e.g. App Service)
2. Select ➡️ Monitoring > Metrics
3. Choose:
   • Signal (e.g. Requests, Response time, 5xx errors)
   • Time range, aggregation, and split by dimension

Pin charts directly to your dashboard or export as alert condition.

Use Case: Tuning App Service Performance

Example signals to monitor in tandem:

• CPU Time (metric)
• Server Errors (5xx) (metric or log)
• Exceptions (App Insights log)
• Outgoing dependencies (App Insights + Log Analytics)

Correlate a performance slowdown by stacking DurationMs, 5xx errors, and infrastructure metrics in the same workbook.

Bicep Add-On: Log Analytics Query Pack (Preview Feature)

Want to standardise performance queries across environments?

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

resource queryPack 'Microsoft.Insights/queryPacks@2022-10-01' = {
  name: 'standard-performance-queries'
  location: resourceGroup().location
}

resource logQuery 'Microsoft.Insights/queryPacks/queries@2022-10-01' = {
  parent: queryPack
  name: 'cpu-usage-by-vm'
  properties: {
    displayName: 'CPU usage by VM (Top Offenders)'
    description: 'Shows VMs with high average CPU usage'
    body: '''
      Perf
      | where ObjectName == "Processor" and CounterName == "% Processor Time"
      | summarize avg(CounterValue) by bin(TimeGenerated, 5m), Computer
      | order by avg_CounterValue desc
    '''
  }
}

Learn More

• Using Metrics Explorer
• Log queries in Azure Monitor
• Performance best practices for App Service