Skip to content

PERFORMANCE_GUIDE.md

Latest commit

 

History

History
139 lines (102 loc) · 3.89 KB

PERFORMANCE_GUIDE.md

File metadata and controls

139 lines (102 loc) · 3.89 KB

Azure CLI Performance Guide

Overview

This guide documents performance characteristics of azlin commands and best practices for minimizing Azure API costs and latency.

Command Performance Matrix

Command Default Latency API Calls Cacheable Optimization
azlin list 0.5-1s 1 (single RG) No Use default RG
azlin list --show-all-vms 10-30s N (all RGs) No Avoid unless needed
azlin list --show-quota +2-8s 1 per region Yes (5min) Use --no-quota for speed
azlin list --show-tmux +1-3s 1 per VM No Use --no-tmux for speed
azlin new 3-5min ~10-15 No Cannot optimize
azlin connect <1s 0 No Already fast

Performance Best Practices

1. Use Configured Resource Group

Fast ✅:

# Set default RG once
echo 'default_resource_group = "my-rg"' >> ~/.azlin/config.toml

# Fast queries forever
azlin list  # 0.5-1s

Slow ❌:

# No configured RG, forced to use --show-all-vms
azlin list --show-all-vms  # 10-30s every time

2. Disable Quota Fetching for Dev/Test

Fast ✅:

azlin list --no-quota  # Skip 2-8s of quota API calls

Use quota only when needed:

azlin list --show-quota  # When capacity planning

3. Disable Tmux Queries for Quick Checks

Fast ✅:

azlin list --no-tmux  # Skip 1-3s of SSH connections

4. Combine Optimizations

Fastest possible list:

azlin list --no-quota --no-tmux  # 0.3-0.5s

Default (balanced):

azlin list  # 2-4s with quota & tmux

API Call Breakdown

azlin list (Default)

  1. az vm list --resource-group <rg> --show-details (0.5-1s)
  2. az network public-ip list --resource-group <rg> (0.3-0.5s, batched)
  3. If --show-quota (default): az vm list-usage --location <region> per region (1-3s each)
  4. If --show-tmux (default): SSH to each running VM (0.5-1s each, parallel max 10)

Total: 2-8s depending on number of VMs and regions

azlin list --show-all-vms

  1. az vm list (NO --resource-group) (10-30s for large subscriptions)
  2. az vm show for each VM to get tags (N API calls)
  3. Same quota & tmux overhead as default

Total: 10-60s depending on subscription size

Quota Fetching Performance

Implementation Details

  • Parallelization: ThreadPoolExecutor with max_workers=10
  • Cache: 5-minute TTL per region:quota pair
  • API: az vm list-usage --location <region>

Cost Analysis

Scenario Regions API Calls Latency Cached Benefit
Single region 1 1 1-2s 5 min reuse
Multi-region 3 3 (parallel) 2-3s 5 min reuse
Many regions 5+ 5+ (parallel, max 10) 3-8s 5 min reuse

When Quota is Useful

  • ✅ Capacity planning near limits
  • ✅ Multi-region deployments
  • ✅ Team quota monitoring
  • ❌ Dev/test with abundant quota (most users)

Recommendation

For most dev/test users (95%+ of quota available):

# Add to your shell alias
alias azl='azlin list --no-quota --no-tmux'  # Fast 0.3-0.5s

Future Optimizations

See open issues:

  • #217: Change --show-quota default to False
  • #218: Remove dead code from vm_manager.py

Measuring Your Performance

# Time a command
time azlin list

# Compare with optimizations
time azlin list --no-quota --no-tmux

Cost Impact

Azure API calls are free (within rate limits), but latency impacts developer productivity:

Daily Lists Default (3s) Optimized (0.5s) Time Saved
10 30s 5s 25s/day
50 150s (2.5min) 25s 125s/day
100 300s (5min) 50s 250s/day (4min)

For teams checking VMs frequently, optimizations save significant time!