Cloud Cost Optimization in Sri Lanka: AWS/Azure Checklist & Case Study
At a Glance - Cloud Cost Optimization Sri Lanka (2026)
- Realistic first-audit savings: Often 20-40% when dev runs 24/7, instances are oversized, and tagging is weak - not a guaranteed 60%
- Start with: Cost Explorer / Azure Cost Management export, 30-day utilization, top 10 services by spend
- Quick wins: Rightsizing, idle resource cleanup, S3 lifecycle, NAT/egress review, dev schedules
- Commitments: Savings Plans / Reserved Instances only after 3-6 months of stable baseline usage
- FinOps: Named owner, mandatory tags, monthly review, budgets at 80/90/100%
- Audit deliverable: Before/after line items with evidence screenshots archived
Introduction
Cloud cost optimization Sri Lanka teams need a repeatable audit - not slogans. Monthly AWS, Azure, or GCP bills often climb because nobody owns the number: dev environments stay on overnight, databases are sized for peak year-round, snapshots accumulate, and data transfer through NAT gateways adds silent margin.
This guide is a practical AWS cost optimization Sri Lanka and multi-cloud checklist: rightsizing, commitments, storage and network savings, tagging and budgets, FinOps Sri Lanka ownership, and a documented before/after audit format we deliver for clients. We do not promise "save up to 60%" - savings depend on how wasteful the starting point is. Well-run accounts may see 10-20%; neglected lift-and-shift stacks often see 25-45% on the first pass without hurting production SLAs.
For migration context, see cloud migration strategy. For ongoing ops, see cloud solutions and DevOps services.
Why Cloud Bills Grow
Before you reduce cloud costs, map waste to root causes:
- Overprovisioning: Production VMs sized for launch-day peak, never reviewed
- Always-on non-prod: Staging and dev running nights and weekends
- Storage drift: Snapshots, logs, and backups with no retention policy
- Network tax: NAT gateway hours, cross-AZ chatter, egress without CDN
- No tags: Cannot attribute spend to team, client, or environment
- No guardrails: Engineers create resources without budget alerts or approval
Cloud is pay-for-what-you-provision, not pay-for-what-you-use, unless you engineer for usage. Treat optimization as a quarterly discipline, not a one-time hack.
Cloud Cost Audit Checklist
Run this checklist in order. Archive Cost Explorer (or Azure Cost Management) screenshots at each step - that is your proof pack.
| Step | Action | AWS | Azure | Done? |
|---|---|---|---|---|
| 1 | Export last 3 months spend by service and linked account | Cost Explorer | Cost Management + Billing | ☐ |
| 2 | List top 10 cost drivers; flag anything >10% of total | Cost Explorer → Group by Service | Cost analysis views | ☐ |
| 3 | Rightsizing review (14-30 day CPU/RAM/network) | Compute Optimizer / Cost Explorer recommendations | Azure Advisor → Cost | ☐ |
| 4 | Delete or stop idle: unattached EBS, old snapshots, unused ELBs/LBs, orphaned IPs | Trusted Advisor / manual inventory | Advisor + resource graph | ☐ |
| 5 | Schedule dev/staging shutdown (business hours only) | Instance Scheduler / Lambda | Start/Stop VMs · Automation accounts | ☐ |
| 6 | Storage lifecycle: S3/Blob tiers, log retention, snapshot max age | S3 Lifecycle · Data Lifecycle Manager | Blob lifecycle management | ☐ |
| 7 | Network review: NAT cost, CloudFront/CDN for static assets, cross-AZ traffic | VPC Flow Logs · Cost by usage type | Azure CDN · egress reports | ☐ |
| 8 | Commitment analysis (only for stable baseline) | Savings Plans / Reserved Instances | Reserved VM · Savings Plan | ☐ |
| 9 | Enforce tags; block creates without Environment + Owner + CostCenter | Organizations tag policies · SCPs | Azure Policy | ☐ |
| 10 | Budgets + anomaly alerts; monthly FinOps review on calendar | AWS Budgets · Cost Anomaly Detection | Cost alerts · budgets | ☐ |
Rightsizing Compute
Rightsizing matches instance size to measured utilization. Review at least 14 days including peak traffic (payday weekends, campaigns, school holidays).
- Export CPU, memory, and network metrics from CloudWatch, Azure Monitor, or GCP Monitoring
- Apply headroom: if peak CPU is 55%, target ~65-70% at peak - not 10% average on a 16 vCPU box
- Prefer newer families where price-performance improved (e.g. AWS Graviton where app supports ARM)
- Migrate EBS gp2 volumes to gp3 (typically lower $/GB with configurable IOPS)
- Re-test after change: latency, error rate, queue depth - cost saved is worthless if SLAs break
Rightsizing is usually the highest-confidence win because evidence comes directly from utilization graphs - not vendor marketing percentages.
Reserved Capacity and Savings Plans
Commitments trade flexibility for discount. AWS and Azure publish discount ranges on their pricing pages (often roughly 30-40% for 1-year, higher for 3-year on steady compute) - your actual savings depend on coverage and payment option. Do not buy 3-year commitments before architecture stabilizes.
| Instrument | Best for | Risk |
|---|---|---|
| AWS Compute Savings Plan | Mixed instance families, evolving app tiers | Commit to $/hour spend; unused commitment is wasted |
| EC2 Reserved Instances | Stable RDS-like workloads, fixed instance type/region | Harder to change shape; use convertible RIs if unsure |
| Azure Reserved VM | .NET stacks on fixed VM SKUs; stacks with Hybrid Benefit if licensed | Same lock-in risk; size reservations after rightsizing |
| GCP Committed Use Discounts | Stable GKE node pools or Compute Engine baselines | Spend-based CUDs more flexible than resource-bound |
Rule of thumb: Cover 50-70% of steady production baseline with 1-year commitments first. Leave burst capacity on-demand or spot for batch/CI workloads.
Storage Savings
Storage looks cheap per GB until snapshots and logs compound.
- S3 / Blob tiers: Standard for hot data; Infrequent Access or Cool after 30-90 days; Glacier for archives with defined retrieval SLA
- Lifecycle policies: Automate transitions and expiry - do not rely on manual cleanups
- Snapshots: Retain 7-30 days for most apps unless compliance requires longer; delete AMI chains nobody owns
- Logs: Ship to cheaper storage with retention limits; avoid infinite CloudWatch log groups
- Intelligent-Tiering: Useful when access patterns are unknown - monitor monitoring cost vs savings
Network and Data Transfer
Network lines on invoices surprise teams who only watched compute:
- NAT gateways: Charge per hour plus per-GB processed - use VPC endpoints for S3/DynamoDB where possible
- CDN: CloudFront, Azure CDN, or Cloud CDN for static assets - reduces origin egress and improves latency to Sri Lankan users via edge POPs
- Cross-AZ traffic: Chatty microservices across AZs add up - colocate hot paths or cache
- Region choice: ap-southeast-1 (Singapore) is typical for Sri Lanka latency - avoid accidental multi-region replication without need
The Ravanaa News migration combined CDN offload with a refactored Next.js stack - origin load dropped sharply, which is both a performance and a cost lever.
Tagging, Budgets and Alerts
You cannot optimize what you cannot attribute. Minimum tag set:
Environment- production, staging, devOwnerorTeam- engineering, data, client nameCostCenter- finance code for chargebackApplication- service or product name
Configure monthly budgets with email alerts at 80%, 90%, and 100% of forecasted spend. Enable cost anomaly detection where available. For IaC teams, add Infracost to pull requests so engineers see $ impact before merge.
FinOps Ownership
FinOps Sri Lanka teams succeed when one named person owns the monthly review - not "everyone" and not "finance alone."
| Role | Responsibility |
|---|---|
| FinOps lead (often CTO or senior dev) | Monthly review, budget vs actual, initiative tracker |
| Engineering | Rightsizing changes, schedules, architecture fixes |
| Finance | Chargeback/showback reports, commitment approvals |
| MSP / partner (optional) | Audit execution, IaC guardrails, managed patching - cloud solutions |
Monthly FinOps meeting (30 min): (1) spend vs budget, (2) top 3 spikes explained, (3) open optimization tickets, (4) commitment coverage check. Document decisions in a shared log - that log becomes your audit trail.
Related reading: cloud solutions Sri Lanka · DevOps CI/CD · cloud security checklist.
Need a documented cloud cost audit?
Hashtag Coders delivers AWS and Azure cost reviews with before/after line items, rightsizing plans, tagging policies, and FinOps setup - from Jaffna for Sri Lankan and international clients.
Documented Before/After: Audit Worksheet Example
This is the deliverable format we use after a 2-week cloud cost optimization Sri Lanka audit on AWS (ap-southeast-1). Line items are an illustrative worked example - replace figures with your own Cost Explorer export. Totals must reconcile before/after; archive CSVs and change tickets as evidence.
| Category | Before (USD/mo) | Action taken | After (USD/mo) |
|---|---|---|---|
| EC2 (app + workers) | $1,140 | Rightsized 6 instances; dev/staging schedule off 7pm-7am | $710 |
| RDS PostgreSQL | $420 | Downsized staging DB; 1-year RI on production | $295 |
| S3 + snapshots | $310 | Lifecycle to IA/Glacier; deleted 180-day-old snapshots | $125 |
| NAT + data transfer | $380 | S3 gateway endpoint; CloudFront for static assets | $240 |
| Idle resources | $190 | Removed unattached EBS, unused ELB, orphaned Elastic IPs | $0 |
| Total | $2,440 | - | $1,370 |
| Savings | $1,070/month (44%) · ~LKR 342K/month at USD/LKR 320 · Performance SLAs unchanged (p95 API latency within ±5%) | ||
Evidence pack (per engagement): Cost Explorer monthly CSV (before/after), Compute Optimizer recommendations, change tickets per resource, budget alert screenshots. Savings vary by account - the 44% figure above reflects common waste patterns (idle dev, oversized prod, snapshot drift), not a guarantee for your stack.
Architecture case: Ravanaa News (performance + origin load)
After migrating Ravanaa News from legacy hosting to a CDN-backed Next.js stack, mobile LCP improved from 4.8s to 1.6s and Lighthouse performance from 52 to 91 - documented in our migration case study. Origin compute and egress dropped because static assets and cacheable pages served from edge - a recurring cost lever for media and content sites, separate from VM rightsizing.
Ongoing Discipline
Cost optimization is not a project with an end date:
- Quarterly rightsizing review after traffic pattern changes
- Annual commitment renewal only after usage review
- New services must pass tag policy and budget check
- Include cost estimate in architecture docs for new features
- Pair with CI/CD so test environments are ephemeral where possible
Frequently Asked Questions
How much can we realistically save?
First audits on neglected accounts often find 25-45% waste. Accounts already rightsized with commitments in place may see 10-20%. Anyone quoting a flat "60% guaranteed" without reading your Cost Explorer export should be treated skeptically.
AWS or Azure - does optimization differ?
Principles are the same: measure, rightsizing, commitments for baseline, storage lifecycle, network review, tags, budgets. Tool names differ; the audit checklist above maps both.
Should we buy 3-year reserved instances now?
Only if instance families and regions are stable for 12+ months proven in data. Most Sri Lankan SMEs should start with 1-year Savings Plans or convertible RIs at partial coverage.
What does a Hashtag Coders cost audit include?
Spend export analysis, rightsizing and idle resource report, storage and network recommendations, tagging policy draft, budget setup, before/after line-item projection, and a 30-day post-change verification pass. Scoped as fixed-fee discovery - typically 2-3 weeks.
Conclusion
Cloud cost optimization Sri Lanka works when you replace hype with evidence: run the audit checklist, document before/after line items, assign FinOps ownership, and treat savings as a range tied to your starting waste - not a universal 60% claim. Start with rightsizing and idle cleanup (low risk), then commitments for stable baselines, then architecture changes like CDN and serverless where they fit. Cloud solutions from Hashtag Coders includes migration, managed ops, and cost reviews with proof packs finance can audit.