Migrating PHP Services to Golang: Performance Optimization and Concurrency Practices

1. Introduction

PHP offers rapid development for small‑to‑medium web services, but its blocking I/O model leads to high latency when many downstream APIs are called, especially under heavy load, causing poor user experience and stability issues.

1.1 Problems with PHP

The mini‑program front‑end calls up to 11 downstream services per request; in a blocking I/O scenario, delays accumulate, and any failure in a layer reduces the overall success probability P = (1‑P1)*(1‑P2)*…*(1‑PN). Adding cache can reduce backend calls but sacrifices real‑time freshness.

1.2 Why Choose Golang

Golang provides lightweight goroutine concurrency, multiplexing, and channels, enabling parallelism at the module level, asynchronous I/O, and a powerful standard library that can replace PHP logic while remaining stable and easy to test.

1.3 Golang Performance Advantages

Compact variable storage reduces memory footprint and improves CPU cache efficiency.

Function inlining reduces call overhead for small functions.

Escape analysis‑driven garbage collection moves many allocations to the stack.

Lightweight goroutine scheduling minimizes context‑switch latency.

Efficient stack management keeps goroutine stacks small.

2. Practice

2.1 Goal

Targeted eight second‑hand‑business APIs for refactoring, with a team of three developers over six weeks, writing roughly 20,000 lines of code.

2.2 Implementation Steps

Without an existing framework, the team built a custom Golang service using go mod, go proxy, and the Gin web framework. Key practices included:

Encapsulating goroutine parallelism.

Environment‑specific timeout handling.

Signature verification.

Log level segregation.

Monitoring and alerting (goroutine count, memory).

Unit testing with go test .

Local caching, configuration management, CI/CD via Jenkins, hot‑reloading with gowatch, and service keep‑alive via systemd.

The first interface (nearby) was completed in ten person‑days, with 80% of the listed features implemented alongside it.

2.3 Parallelism in Critical Paths

Two layers of logic in the nearby interface were parallelized (e.g., getErshouList, getNewfangList, getZufangList) and in the search flow (SearchResblockHouseSell, GetResblockSell), reducing latency by over 100 ms. A helper GoWait method simplified concurrent calls.

2.4 Caching Decision

Redis cache was added to the listing page to balance freshness and stability: detail pages remain real‑time, while list pages benefit from cache without causing noticeable inconsistency.

2.5 Performance Monitoring

Golang’s runtime makes it easy to collect goroutine counts and memory usage. Fast‑eye dashboards were used to track trends, set thresholds, and detect leaks or panics early.

2.6 Supporting Tools

go generate – code generation

go fmt – formatting

go mod tidy/vendor – dependency management

go vet, golint – static analysis

goswagger – API documentation

gowatch – hot recompilation

These tools together provided a smooth development experience comparable to PHP.

2.7 Gradual Rollout

Features were released via gray‑scale deployment using Apollo for offline configuration, allowing domain switching and fallback to the old service if needed.

3. Results

Stability improved by ~0.02%.

Average latency reduced by 15.6%.

TP90 latency reduced by 37.5%.

These gains were achieved without a full‑scale Redis cache, relying on concurrency and selective caching. The service has run stably for eight months with continuous iteration.

4. Future Plans

Adopt the company‑wide framework for component standardization and expand Golang usage across more projects.

5. Conclusion

Golang, when applied thoughtfully, delivers significant performance and stability benefits for web services without adding undue complexity, making it a strong candidate for future backend development.