Optimizing Apollo Configuration Reading in PHP: Reducing CPU and Memory Overhead

Background: As the number of cities supported by the Beike app grows, each city’s configuration is stored in Apollo (a distributed configuration center). Reading the entire configuration for every request, especially with around 1,000 cities and dozens of items per city, leads to high CPU and memory usage, causing service degradation under heavy concurrency.

Analysis: The configuration read‑process includes fetching the JSON‑encoded data from Apollo, decoding it, and using it throughout the request. Repeated reads and full JSON decoding create a performance bottleneck.

Optimization approaches:

Avoid repeated reads and perform minimal processing.

Improve read speed.

Solutions implemented:

3.3.1 Use static variables to cache configuration

By decoding the configuration once per request and storing the result in a static variable, subsequent calls reuse the already‑decoded data, eliminating extra Apollo reads and JSON decoding.

Advantages: Single read and decode per request; reduced CPU usage. Disadvantages: Configuration updates are not visible until the request ends.

3.3.2 Minimal json_decode

Only decode the second‑level array for the needed city_id, avoiding unnecessary decoding of unrelated entries.

Advantages: Faster when only a subset of data is needed. Disadvantages: No benefit when the target key size approaches the total size.

3.4.1 Use Yac or APCu as local cache

Yac (lock‑free shared memory) and APCu (user data cache) store the configuration locally, providing read speeds comparable to memcached with lower overhead.

Advantages: Memory usage lower than Apollo + JSON; execution time about one‑quarter of the original method. Disadvantages: Requires additional extensions and a daemon to keep the cache synchronized with Apollo.

3.4.2 Reverse‑engineer configuration into PHP files

Export configuration to PHP files using var_export , then load them with require and benefit from OPcache for fast reads.

Advantages: Significant read‑time improvement. Disadvantages: Adds file‑maintenance overhead; real‑time updates require redeployment or submodule updates.

3.5.1 Use MessagePack format

Store configuration in MessagePack, a binary serialization format that is smaller and faster to parse than JSON.

Advantages: About 60% of JSON parsing time with reduced memory footprint. Disadvantages: Introduces a new PHP extension dependency.

Conclusion: Combining Apollo with static variables, Yac/APCu caching, and selective decoding dramatically reduces CPU and memory consumption, preventing service outages. Depending on the scenario, these techniques can be mixed to achieve the best performance while balancing complexity and freshness requirements.