Optimizing Apache HttpClient for High-Concurrency Scenarios

1. Background

We have a business that calls an HTTP service provided by another department, with daily call volume at the ten‑million level. Using HttpClient, the original average execution time was 250 ms; after optimization it dropped to 80 ms, eliminating thread‑exhaustion alarms.

2. Analysis

Original implementation created a new HttpClient and HttpPost for each request, then closed response and client explicitly. The following problems were identified:

2.1 Repeated creation of HttpClient

HttpClient is thread‑safe; creating a new instance per request adds unnecessary overhead. A single shared instance should be used.

2.2 Repeated TCP connection establishment

Each request performed a full TCP handshake and teardown, consuming several milliseconds. Using keep‑alive to reuse connections dramatically reduces this cost.

2.3 Redundant entity buffering

Code copied the response entity into a String while the original HttpResponse still held the content, leading to double memory usage and the need for explicit connection closure.

HttpEntity entity = httpResponse.getEntity();

String response = EntityUtils.toString(entity);

3. Implementation

Three main actions were taken: a singleton HttpClient, a connection‑pool manager with keep‑alive, and a better response handling strategy.

3.1 Define a keep‑alive strategy

Custom ConnectionKeepAliveStrategy reads the "timeout" parameter from the response header; if absent, a default of 60 seconds is used.

ConnectionKeepAliveStrategy myStrategy = new ConnectionKeepAliveStrategy() {
    @Override
    public long getKeepAliveDuration(HttpResponse response, HttpContext context) {
        HeaderElementIterator it = new BasicHeaderElementIterator
            (response.headerIterator(HTTP.CONN_KEEP_ALIVE));
        while (it.hasNext()) {
            HeaderElement he = it.nextElement();
            String param = he.getName();
            String value = he.getValue();
            if (value != null && param.equalsIgnoreCase("timeout")) {
                return Long.parseLong(value) * 1000;
            }
        }
        return 60 * 1000; // default 60s
    }
};

3.2 Configure a PoolingHttpClientConnectionManager

Set maximum total connections and per‑route limits according to business needs.

PoolingHttpClientConnectionManager connectionManager = new PoolingHttpClientConnectionManager();
connectionManager.setMaxTotal(500);
connectionManager.setDefaultMaxPerRoute(50); // example values

3.3 Build the HttpClient

httpClient = HttpClients.custom()
        .setConnectionManager(connectionManager)
        .setKeepAliveStrategy(myStrategy)
        .setDefaultRequestConfig(RequestConfig.custom()
                .setStaleConnectionCheckEnabled(true).build())
        .build();

Note: Using setStaleConnectionCheckEnabled is deprecated; a dedicated thread should periodically invoke closeExpiredConnections() and closeIdleConnections() .

public static class IdleConnectionMonitorThread extends Thread {
    private final HttpClientConnectionManager connMgr;
    private volatile boolean shutdown;

    public IdleConnectionMonitorThread(HttpClientConnectionManager connMgr) {
        super();
        this.connMgr = connMgr;
    }

    @Override
    public void run() {
        try {
            while (!shutdown) {
                synchronized (this) {
                    wait(5000);
                    connMgr.closeExpiredConnections();
                    connMgr.closeIdleConnections(30, TimeUnit.SECONDS);
                }
            }
        } catch (InterruptedException ex) {
            // terminate
        }
    }

    public void shutdown() {
        shutdown = true;
        synchronized (this) {
            notifyAll();
        }
    }
}

3.4 Reduce overhead when executing methods

Do not close the connection manually; let the client manage it. Use a ResponseHandler to automatically consume the entity.

public
T execute(final HttpHost target, final HttpRequest request,
        final ResponseHandler
responseHandler, final HttpContext context)
        throws IOException, ClientProtocolException {
    Args.notNull(responseHandler, "Response handler");
    final HttpResponse response = execute(target, request, context);
    try {
        return responseHandler.handleResponse(response);
    } finally {
        final HttpEntity entity = response.getEntity();
        EntityUtils.consume(entity);
    }
}

4. Additional Settings

4.1 Timeout configuration

Configure connection timeout, socket timeout, and connection‑manager timeout, and optionally disable retries.

HttpParams params = new BasicHttpParams();
int CONNECTION_TIMEOUT = 2 * 1000;
int SO_TIMEOUT = 2 * 1000;
long CONN_MANAGER_TIMEOUT = 500L;

params.setIntParameter(CoreConnectionPNames.CONNECTION_TIMEOUT, CONNECTION_TIMEOUT);
params.setIntParameter(CoreConnectionPNames.SO_TIMEOUT, SO_TIMEOUT);
params.setLongParameter(ClientPNames.CONN_MANAGER_TIMEOUT, CONN_MANAGER_TIMEOUT);
params.setBooleanParameter(CoreConnectionPNames.STALE_CONNECTION_CHECK, true);
httpClient.setHttpRequestRetryHandler(new DefaultHttpRequestRetryHandler(0, false));

4.2 Nginx keep‑alive

If an Nginx reverse proxy is used, configure keepalive_timeout , keepalive_requests , and upstream keepalive to match the client settings.

After applying these changes, the average request latency decreased from 250 ms to roughly 80 ms, demonstrating a substantial performance gain.