Site Reliability Engineering (SRE), SLI, SLO, SLA & Error Budgets

Lab 19 of 20 · SRE & Error Budgets · 5 hands-on activities · bilingual (English + සිංහල)

පිටුව 19: Google SRE විප්ලවය, විශ්වසනීයත්වයේ ගණිතය: SLI, SLO, SLA සහ Error Budgets

Google හි Ben Treynor Sloss, SLI (මනින දේ), SLO (ඉලක්කය), SLA (ගිවිසුම), Error Budget පිච්චෙන වේගය (Burn Rate Graph) සහ Feature Freeze.

Architectural Foundation & Systems Diagram

Site Reliability Engineering was created at Google in 2003 by Ben Treynor Sloss: "SRE is what happens when you ask a software engineer to design an operations team."

• 📊 SLI (Indicator) — What is measured. Ratio of good events to total valid events:
  SLI = (Good Requests / Total Requests) * 100%

• 🎯 SLO (Objective) — Internal Engineering Target. The reliability goal agreed upon with product managers (e.g. 99.9% uptime over rolling 30 days).

• 📜 SLA (Agreement) — Legal Contract with Customers. Carries financial penalties if breached. SLO must ALWAYS be stricter than SLA!

Step-by-step practical lab

Step 01: Calculate Real-Time SLI Availability Percentage

සිංහලෙන් — සැබෑ දත්ත අනුව SLI Availability ප්‍රතිශතය ගණනය කරමු.

Calculate the Service Level Indicator: (Good Requests / Total Requests) * 100%.

python3 -c "good=999800; total=1000000; sli=(good/total)*100; print(f'SLI Availability: {sli:.3f}%')"

• SLI formula : (Good Requests / Total Requests) * 100% = සැබෑ පද්ධති සාර්ථකත්ව අනුපාතය.

Expected output

SLI Availability: 99.980%

Step 02: Calculate 30-Day Error Budget Allowance

සිංහලෙන් — 99.9% Uptime එකකදී අපට හිමි Downtime විනාඩි ගණන ගණනය කරමු.

Calculate allowed monthly downtime for a 99.9% Service Level Objective (SLO).

python3 -c "mins = 30*24*60; slo = 0.999; allowed = mins*(1-slo); print(f'Allowed Downtime: {allowed:.2f} minutes/month')"

• -c : Python script එක කෙලින්ම CLI එකෙන් run කරයි.

Expected output

Allowed Downtime: 43.20 minutes/month

Step 03: Evaluate 14.4x Critical Burn Rate Alert Threshold

සිංහලෙන් — Error Budget එක විනාශ වන වේගය (Burn Rate) ගණනය කරමු.

Calculate the current 1-hour burn rate multiple given observed error rates.

python3 -c "errors=48; requests=10000; err_rate=errors/requests; slo=0.001; burn=err_rate/slo; print(f'Burn Rate: {burn:.1f}x (Threshold: 14.4x)')"

• burn rate : Burn rate = observed error rate / allowed error rate.

Expected output

Burn Rate: 4.8x (Threshold: 14.4x)

Hands-on practice activities

Work through these yourself — tick them off as you go.

• Activity 91: Calculate SLI Availability Percentage (Checkpoint) — If 999,000 requests succeeded out of 1,000,000, calculate: SLI = (999,000 / 1,000,000) * 100% = 99.9%.
• Activity 92: Calculate 30-Day Three Nines (99.9%) Downtime (Checkpoint) — Understand that 99.9% availability permits exactly ~43.2 minutes of total allowable downtime per month.
• Activity 93: Evaluate 14.4x Burn Rate Alert (Checkpoint) — A 14.4x burn rate burns 100% of your 30-day error budget in just 48 hours (2 days), requiring an immediate SEV-1 page!
• Activity 94: Enforce Release Freeze Rule (Checkpoint) — When the error budget is exhausted (0% remaining), freeze all new feature deployments and focus 100% on reliability!
• Activity 95: SRE Origins Quiz (Quiz) — Who founded Site Reliability Engineering at Google in 2003? Answer: Ben Treynor Sloss.

Keep going: ← Lab 18 · Lab 20 → · Track your progress in the Self Guide