Understanding DBLE Network Module: TCP/IP Stack, MySQL Protocol, and BIO vs NIO

Preface

For computer science, knowledge of computer networking is essential. While many developers focus on high‑level frameworks, low‑level network I/O handling is crucial for middleware and framework developers. DBLE's network module is written entirely with native JDK code, and studying its source helps understand high‑performance network I/O.

This article is the first in the "DBLE Network Module Source Code Analysis" series and explains basic network I/O concepts.

1. TCP/IP Protocol Stack

The TCP/IP protocol simplifies the OSI model’s seven layers into four: link, network, transport, and application. Data is encapsulated layer by layer, as illustrated below.

Each upper‑layer protocol’s data becomes the payload for the next lower layer, forming a layered encapsulation structure.

2. MySQL Protocol

The MySQL protocol operates at the application layer of the TCP/IP stack, meaning its data is carried as the TCP payload. Because TCP is a stream‑oriented protocol, it suffers from packet‑sticking (粘包) and splitting (拆包) issues that must be resolved at the application layer.

Common solutions include:

Fixed‑length messages.

Delimiter characters (e.g., carriage return).

Message headers that specify total length.

More complex application‑layer protocols.

MySQL solves the problem by using fixed‑length messages combined with a length field in the header, as shown below.

https://dev.mysql.com/doc/dev/mysql-server/latest/page_protocol_basic_packets.html

Since DBLE can connect like a MySQL client, it must be able to parse MySQL packets; the handling logic will be explored in later source‑code analyses.

3. BIO and NIO

DBLE achieves high performance through high‑efficiency network I/O. It uses NIO (non‑blocking I/O) with I/O multiplexing for both front‑end client connections and back‑end MySQL connections.

When connection counts are low, a BIO + multi‑threaded approach can be fast, but as connections grow, thread‑switching overhead outweighs the benefits, degrading performance.

NIO, by contrast, allows a small number of threads to manage many connections without creating a dedicated thread per connection, and combined with asynchronous request processing, it forms the secret of DBLE’s high performance.

Next Issue Preview

This article covered basic network I/O concepts—TCP/IP stack, MySQL protocol, and BIO/NIO—preparing readers for the next piece, which will dive into DBLE’s actual network handling source code.