How Huolala Cut Cloud Costs and Boost Efficiency with FinOps

Brief on Cloud Cost Issues – Reducing Costs and Increasing Efficiency

In 2021 the company highlighted blind spots in technical spending and adopted the slogan “eliminate blind spots, find breakthroughs, reduce costs and increase efficiency,” making cost reduction the top priority for 2022.

The CTO emphasized that stability is the foundation for all technical success; without stable technology, other efforts are nullified.

Efficiency, stability, and cost often conflict under current technical means, so balancing them became a key 2022 challenge.

Huolala's IT Infrastructure Load

2.1 Geographic Distribution of Infrastructure

The infrastructure spans multiple cloud providers (AWS, Azure, Alibaba Cloud, Huawei Cloud, Tencent Cloud) and private data centers across regions such as Singapore, Brazil, Guangzhou, Shenzhen, etc., supporting services like servers, databases, load balancers, storage, containers, logging, and security components at a ten‑thousand‑level scale for core links and big‑data workloads.

Cost governance is complicated by the diversity of providers, locations, service types, scale, and business domains.

2.2 Historical Cost Trend (HLLM)

The chart shows that before optimization costs rose parabolically, while after optimization the cost curve slopes downward.

IT Cost Composition and Optimization Recommendations

IT cost = Service usage cost (S) + Labor cost (H).

Service cost factors include infrastructure cost (B), variable cost from service selection (X), utilization (U), business scale (Y), and growth factor (N). Optimizing service utilization (U) and selecting cost‑effective services (X) are the primary levers.

Labor cost H = Development cost (D) + Maintenance cost (M). Improving development efficiency (DE) and operations efficiency (ME) reduces labor expenses.

Key Components and Work of Cost Optimization

The Cloud Resource Delivery team, part of operations, drives cost‑saving initiatives based on FinOps practices.

4.1 Controlling Service‑Selection Variable X

Examples of concrete optimizations:

Storage type: migrate from older disks (gp2) to newer (gp3) or replace SSD with HDD where appropriate, achieving ~20% cost reduction.

Network design: use gateway endpoints instead of NAT/public access for S3, saving tens of thousands of dollars per month; similar strategies for Azure Blob and cross‑region S3 access.

Logging system: replace self‑built log clusters with OpenSearch and object‑storage‑backed warm data, and switch to Graviton2 ARM instances, cutting costs to 40% of the original.

Spot instances: achieve >90% coverage in test environments and apply to non‑core production workloads.

Containerization: replace traditional ECS with containers, raising average utilization from 30% to 50% and lowering costs by ~14%.

Reserved instances / Savings Plans: maintain ~95% coverage and near‑100% utilization, integrating purchasing decisions into the cloud‑management platform.

Disaster‑recovery: conduct DR drills costing ~1.5% of monthly spend and backups ~1%.

4.2 Cloud Management Platform – Lcloud

Lcloud, developed by the Cloud Platform team, provides two main capabilities:

Cost reduction: unified visibility and optimization of cloud and third‑party service expenses.

Efficiency increase: streamlined resource delivery and automated cost‑governance workflows.

The platform supports multi‑dimensional chargeback, cost‑autonomy alerts, accurate cost forecasting (within 5% error), integration with other internal tools, and one‑stop operations management.

4.3 Cost Governance Committee

Service owners from each department act as cost interface contacts.

Operations engineers treat cost as a KPI, monitoring it alongside performance metrics.

Monthly cost‑review meetings set KPIs and track optimization progress.

Why Cloud Costs Often Spike

5.1 Cost Blind Spots

Lack of deep knowledge about cloud services leads to explosive cost growth. Common issues include improper network configuration (NAT or private link usage), inappropriate storage tier selection, outdated disk types, and un‑evaluated managed services.

Mitigation recommendations:

Consult experts for usage‑based services.

Align capacity provisioning with actual SLA and performance needs; avoid over‑provisioning.

Maintain comprehensive cost monitoring to catch unexpected spend early.

5.2 Stage‑Based Cost Inputs

Cost evolves with business stages, from initial decline to gradual increase as stability becomes the primary concern and later as cost governance adds planned stability spend.

5.3 Elasticity

Elasticity is a core cloud benefit but can drive cost up if resource usage is not accurately predicted or if discount opportunities are missed.

Exploring Business Value of Cost Governance

FinOps combines finance, engineering, and business to enable data‑driven spending decisions. Huolala’s cost‑optimization maturity surpasses 95% in cost allocation, 90%+ discount coverage, and keeps budget forecast error within 5%.

Accurate cost insight allows the company to align activities with business value and continue advancing cost‑business metric relationships.