Analyzing MySQL Connection Latency in Java Web Applications

The article explores the overhead of creating a MySQL connection in a Java web application, aiming to quantify the time spent in each phase of the TCP‑based MySQL handshake and authentication process.

It begins with a minimal Java snippet that loads the MySQL driver and opens a connection:

Class.forName("com.mysql.jdbc.Driver");

String name = "xttblog2";
String password = "123456";
String url = "jdbc:mysql://172.16.100.131:3306/xttblog2";
Connection conn = DriverManager.getConnection(url, name, password);
// program ends, connection is forcibly closed

Using Wireshark, the author captures the full handshake, which consists of three main steps:

Step 1: Establish a TCP connection via the three‑way handshake.

Step 2: Server sends a handshake packet; the client responds.

Step 3: Client sends an authentication packet; upon successful verification the server returns an OK packet and the session is ready for commands.

The capture shows that the entire connection establishment takes approximately 225 ms (10.416042 s – 10.190799 s), which includes at least seven round‑trips between client and server.

When the connection is properly closed with Connection.close() , the shutdown sequence is captured as a four‑step TCP termination, taking about 183 ms:

Class.forName("com.mysql.jdbc.Driver");

String name = "shine_user";
String password = "123";
String url = "jdbc:mysql://172.16.100.131:3306/clever_mg_test";
Connection conn = DriverManager.getConnection(url, name, password);
conn.close();

Extrapolating these numbers to a site with 100 000 requests per day (assuming 150 ms per connection) shows that merely establishing connections would consume roughly 4 hours of CPU time daily, highlighting the critical need for a connection pool.

The article concludes that database connections are expensive, and recommends using connection pooling together with caching, SQL pre‑compilation, and load balancing to mitigate the performance impact.