Measuring Heap Memory Allocation per HTTP Request in SpringBoot

Introduction: The author, a senior architect, explains the need to quantify heap memory usage per RPC/HTTP request to optimize GC and performance.

Experiment design: Create a SpringBoot 2.5.4 application with a POST endpoint and a GET endpoint to trigger GC; use JMeter with 10 threads each performing 2000 requests (total 20,000). Enable detailed GC logging and set JVM options (e.g., -Xmx4g -Xms4g -XX:SurvivorRatio=8 -Xmn2g, etc.).

Code snippet for the controller:

@Slf4j
@RestController
public class TestController {
    private AtomicLong count = new AtomicLong(0);
    @ResponseBody
    @RequestMapping(value = "create", method = RequestMethod.POST)
    public String create(@RequestBody Order order) {
        //log.warn("收到提单请求 cnt{}:{}", count.getAndIncrement(), order);
        return "ok";
    }
    @ResponseBody
    @RequestMapping(value = "gc", method = RequestMethod.GET)
    public String gc() {
        System.gc();
        return "ok";
    }
}

JMeter configuration: set thread group, HTTP defaults, request headers (Content-Type: application/json), and request body.

Execution: Run JMeter, manually trigger GC before the test, then collect GC logs to calculate new generation heap growth.

Results: Average heap allocation per HTTP request is about 34 KB even with a tiny 50‑character payload; adding a larger payload (1200 characters) increases it to ~36 KB. Logging the request adds ~20 KB, raising usage to ~56 KB. Removing the payload drops usage to ~35.7 KB.

Analysis: The memory cost is dominated by SpringBoot’s internal object creation rather than the request body size; excessive logging significantly increases memory and GC frequency.

Conclusion: Controlling log size and understanding per‑request memory allocation are crucial for backend performance; in larger production systems, RPC requests can consume 0.5–1 MB, requiring multiple young GCs per minute under high concurrency.