Technical Cost Optimization and Fine‑Grained Operations: Strategies, Processes, and Best Practices

1 Cost Optimization

Cost optimization is a "stop‑bleeding" process that addresses hidden waste in technology spending, especially during rapid business growth.

1.1 Build the Team

A dedicated team is essential for cross‑departmental cost‑saving initiatives. Typical roles include:

Owner (Team Lead) : overall planning, scheduling, coordination, and external reporting.

Ops Lead (Ops Director / SRE Lead) : evaluate and optimize overall technical costs, public‑service costs, and cloud‑provider operations.

Business Tech Lead (Business Tech Director / Manager) : assess and implement cost optimizations for specific business units.

Finance Lead (Finance Director / Manager) : confirm, verify, and budget technical cost expenditures.

1.2 Current Situation Analysis

Technical cost waste often stems from a lack of visibility—costs are a "black box". The analysis should first be quantitative (cost amount, instance count) and then qualitative.

Two dimensions are recommended: time‑based (monthly, yearly) and product‑based (per‑cloud‑product usage).

If reporting tools are insufficient, a simple Excel sheet can be used:

Month

Cloud Vendor 1 Cost

…

Total

2021-01

Alibaba Cloud

…

Total

Visualize trends with line charts and composition with pie charts, using tools like Feishu multi‑dimensional tables or Excel.

1.2.1 Discounts and Billing Modes

Negotiate volume discounts with cloud vendors; different products may have separate discount opportunities. Billing modes (annual vs. monthly) also affect cost.

1.2.2 Storage

Historical storage : unused temporary or expired files.

Useless storage : old buckets left after new ones are created.

Cold/HOT storage : infrequently accessed data kept in standard storage instead of low‑frequency or archival tiers.

Log storage : long‑term retention of trace logs that lose value quickly; consider sampling or cleaning after a few hours.

Storage bandwidth : evaluate CDN + origin vs. direct external access.

Storage compute : image processing or other compute features may incur extra fees; assess trade‑offs.

1.2.3 Machines

Idle machines : resources left running after scheduled tasks end.

Low‑load production machines : over‑provisioned instances that are rarely used.

Idle non‑production machines : test or dev environments with high specs but low usage.

Database waste : stale data kept after migrations or version upgrades.

1.2.4 Traffic Bandwidth

Internal vs. external traffic : avoid routing internal traffic through external networks.

Billing mode : compare bandwidth‑based vs. traffic‑based pricing; choose the most cost‑effective model.

1.2.5 CDN

CDN bandwidth : enable compression, especially for images, to halve bandwidth usage.

Other CDN costs : HTTPS fees vary by vendor; negotiate discounts where possible.

1.2.6 Cloud Services

Service sharing : consolidate DDoS or security services across multiple businesses.

Idle services : discontinue unused governance or auxiliary services.

Version differences : newer service versions may have different pricing.

Over‑provisioned services : downgrade configurations after business review.

Multi‑vendor comparison : evaluate cost and reliability across vendors for common services like content moderation or SMS.

1.3 Project Plan and Milestones

Prioritize items based on expected savings, implementation difficulty, and risk. Three major milestones:

Stop‑bleeding : quick wins to eliminate obvious waste, targeted for completion by month‑end.

Business Optimization : higher‑impact, cross‑team initiatives requiring business involvement, 1‑3 months.

Fine‑grained Operations : establish a sustainable cost‑control system for ongoing adjustments.

Example detailed plan table:

Business

Optimization Measure

Problem Description

Estimated Savings

Current Status

Planned Completion

Owner

Related Docs

Infrastructure

SLB billing review & adjustment

Unreasonable SLB billing

5,000

Planned

2022-11-10

Zhang San

Doc XXX

1.4 Communication and Reporting

Regular reports and meetings keep stakeholders aligned:

Monthly analysis report : cost breakdown, plan execution review, variance analysis.

Business‑tech cost report : adds business‑level cost view for fine‑grained allocation.

Finance analysis report : non‑technical perspective on cost distribution.

Monthly cost‑optimization meetings involve all team members, followed by email/IM distribution to executives (CTO, CFO, etc.).

1.5 Other Principles and Considerations

Do not impact core business operations.

Accept minor efficiency loss for aggressive cost cuts.

First stop‑bleed, then optimize .

Some optimizations have delayed effect (e.g., monthly‑billed CDN).

Prioritize high‑impact items (e.g., storage).

Break down each cost line for evaluation.

Involve business and finance to share responsibility.

Prefer large vendors for reliability and support.

When migrating between cloud providers, negotiate migration cost sharing.

2 Fine‑Grained Operations

Unlike short‑term optimization, fine‑grained operations focus on sustainable, long‑term cost control.

2.1 Organizational Division

2.1.1 Cost Sharing by Business

Shift cost ownership to each business unit; the technical cost‑optimization team coordinates and validates recommendations.

2.1.2 Business‑Driven Cost Strategies

Examples include limiting WeChat "favorites" storage and enforcing retention policies for temporary file‑sharing services, both reducing unnecessary storage consumption.

2.2 Process Mechanisms

2.2.1 Resource Request Process

Introduce a standardized request template (reason, business, specs, cost estimate) and require technical‑cost team approval before provisioning.

2.2.2 Cost Pre‑Estimation Mechanism

At year‑end, forecast next‑year costs using historical data and business targets. Separate fixed costs (e.g., security infrastructure) from variable costs (machines, bandwidth, storage). Common metrics include per‑user storage cost, per‑active‑user CDN cost, per‑active‑user bandwidth cost, per‑active‑user machine cost, and per‑active‑user traffic cost.

2.3 System Support

Initially, manual reporting suffices; later, automate daily/weekly/monthly cost dashboards, alerts, and utilization monitoring (CPU, memory, instance usage) to maintain transparency and rapid response.

3 Conclusion

Technical cost optimization is an ongoing effort that requires continuous iteration, fine‑grained operations, and eventual automation, visualization, and intelligent tooling.

Hi, I am Pan Jin, with over 10 years of R&D management and architecture experience, author, entrepreneur, former Tencent and listed‑company staff, now leading tech management at a C‑round startup. I have a background in NOI and ACM, and stay interested in frontend, cross‑platform, backend, cloud‑native, DevOps, etc. Feel free to connect via WeChat public account "Architecture and the Horizon" or my blog www.phppan.com.