Erlang OTP: Distributed & Fault-Tolerant Systems Programming · 课时

Erlang OTP 案例研究

分析大规模 Erlang OTP 部署中的真实案例和最佳实践，从业界成功经验中学习。

Erlang OTP 案例研究 是 CoddyKit 上的免费 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课时。这是第 3 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Learning from Erlang's Giants

Why is Erlang chosen for mission-critical systems? Its unique strengths – concurrency, fault tolerance, and distribution – make it ideal for applications needing near-perfect uptime.

Today, we'll explore how major real-world projects leverage these Erlang/OTP features to build highly robust and scalable systems.

Ericsson AXD 301: Telecom Reliability

Ericsson's AXD 301, a massive ATM switch, was one of Erlang's earliest and most famous success stories. It achieved "five nines" (99.999%) availability, meaning less than 5 minutes of downtime per year.

This incredible reliability was largely due to Erlang's:

Fault Tolerance: Supervisors automatically restarting failed components.
Hot Code Upgrades: Updating software without service interruption.
Process Isolation: Failures in one part don't bring down the whole system.

Simulating AXD 301's Resilience

The AXD 301's resilience came from processes that could fail and be restarted by supervisors. This tiny example shows a worker that intentionally crashes, and its supervisor immediately restarts it, demonstrating a core Ericsson principle.

-module(crash_demo).
-behaviour(gen_server).
-export([start_link/0, init/1, handle_call/3, handle_cast/2, handle_info/2, terminate/2, code_change/3]).
-export([crash_me/0, main/0]).

%% Worker functions (behaves as a gen_server)
start_link() -> gen_server:start_link({local, ?MODULE}, ?MODULE, [], []).
init([]) -> io:format("Worker started!~n"), {ok, nil}.
crash_me() -> gen_server:call(?MODULE, crash).
handle_call(crash, _From, State) -> 
    io:format("Worker intentionally crashing!~n"), 
    exit(i_crashed), %% Simulate a crash
    {reply, ok, State};
handle_call(_Req, _From, State) -> {reply, ok, State}.
handle_cast(_Msg, State) -> {noreply, State}.
handle_info(_Info, State) -> {noreply, State}.
terminate(_Reason, _State) -> io:format("Worker terminating.~n").
code_change(_OldVsn, State, _Extra) -> {ok, State}.

%% Supervisor part (embedded for this demo)
start_supervisor() ->
    supervisor:start_link({local, demo_sup}, ?MODULE, supervisor). %% Pass 'supervisor' as InitArgs

%% This init/1 is for the supervisor behavior callback
init(supervisor) ->
    SupFlags = #{strategy => one_for_one, intensity => 1, period => 5},
    ChildSpecs = [
        #{id => my_worker,
          start => {?MODULE, start_link, []}, %% Start the worker part of this module
          restart => permanent,
          shutdown => 5000,
          type => worker,
          modules => [?MODULE]}
    ],
    {ok, {SupFlags, ChildSpecs}}.

%% Main entry point for runnable
main() ->
    io:format("Starting supervisor and worker...~n"),
    {ok, _SupPid} = start_supervisor(),
    timer:sleep(100), %% Give worker time to start
    io:format("Worker PID before crash: ~p~n", [whereis(?MODULE)]),
    crash_me(), %% Trigger crash of the worker part
    timer:sleep(100), %% Give supervisor time to restart
    io:format("Worker PID after restart: ~p~n", [whereis(?MODULE)]),
    ok.

WhatsApp: Billions of Messages

WhatsApp handled billions of messages daily with a relatively small engineering team, largely thanks to Erlang. Its architecture efficiently managed massive concurrent user connections by leveraging:

Massive Concurrency: Erlang's lightweight processes (millions per node) allowed handling countless simultaneous users.
Message Passing: Asynchronous message passing between processes mimicked the real-world communication flow.
Distribution: Erlang's built-in distribution enabled seamless scaling across multiple server nodes.

WhatsApp's Core: Simple Messaging

At its heart, WhatsApp is about processes sending messages. This snippet shows two processes communicating, illustrating the fundamental building block of their system.

-module(messenger).
-export([start_sender/1, start_receiver/0, main/0]).

%% Receiver process
start_receiver() ->
    spawn(fun() -> receiver_loop() end).

receiver_loop() ->
    receive
        {message, From, Msg} ->
            io:format("Receiver (~p) got: ~s from ~p~n", [self(), Msg, From]),
            From ! {ack, self()},
            receiver_loop();
        _ ->
            io:format("Receiver got unknown message.~n"),
            receiver_loop()
    end.

%% Sender process
start_sender(ReceiverPid) ->
    spawn(fun() -> sender_loop(ReceiverPid) end).

sender_loop(ReceiverPid) ->
    Msg = "Hello from sender!",
    io:format("Sender (~p) sending '~s' to ~p~n", [self(), Msg, ReceiverPid]),
    ReceiverPid ! {message, self(), Msg},
    receive
        {ack, _Receiver} ->
            io:format("Sender (~p) received acknowledgement.~n", [self()]);
        _ ->
            io:format("Sender got unexpected reply.~n")
    end,
    timer:sleep(100),
    ok. %% Only send one message for this demo

%% Main entry point for runnable
main() ->
    io:format("Starting messaging demo...~n"),
    Receiver = start_receiver(),
    timer:sleep(50), %% Give receiver a moment to start
    Sender = start_sender(Receiver),
    io:format("Sender PID: ~p, Receiver PID: ~p~n", [Sender, Receiver]),
    timer:sleep(500), %% Allow messages to exchange
    ok.

RabbitMQ: Reliable Message Queues

RabbitMQ, a widely used open-source message broker, relies heavily on Erlang/OTP for its robustness and scalability. It provides critical features such as:

Reliability: Persistent message queues ensure messages aren't lost even if the server crashes.
Clustering: Multiple RabbitMQ nodes can form a cluster, sharing queues and data, thanks to Erlang's distribution.
Fault Tolerance: Supervisors manage internal components, ensuring continuous operation and automatic recovery.

Best Practice: Embrace 'Crash First'

A key takeaway from these case studies is Erlang's "crash first" philosophy. Instead of trying to prevent every error, systems are designed to crash cleanly and be restarted by a supervisor. This approach leads to:

Simpler error handling logic.
More robust systems that automatically recover.
Easier identification of root causes through crash reports.

Best Practice: Seamless Hot Upgrades

Another powerful feature utilized in high-availability systems like Ericsson's is hot code loading and upgrades. Erlang allows you to replace running code modules without stopping the application or losing its state.

Essential for systems requiring continuous uptime.
Minimizes maintenance windows.
Enables rapid deployment of fixes and new features.

Best Practice: Scale with Distribution

Erlang's built-in support for distributed computing is fundamental to scaling systems like WhatsApp and RabbitMQ. It allows applications to seamlessly span multiple machines or nodes.

Node Communication: Processes on different machines can communicate as if they were local.
Global Registration: Register process names globally, making them discoverable across the cluster.
Fault Tolerance: Distribute workload and ensure that the failure of one node doesn't bring down the entire system.

Check Your Understanding

Based on the real-world case studies discussed, which of the following are key benefits of using Erlang/OTP for building highly available and scalable systems?

Recap: Learning from Success

We've explored how major projects like Ericsson AXD 301, WhatsApp, and RabbitMQ leverage Erlang/OTP's unique strengths.

Fault tolerance via supervision allows systems to recover automatically from failures.
Hot code upgrades enable continuous service without downtime, crucial for critical systems.
Massive concurrency and distribution are key to scaling applications and building resilient architectures.

These principles are central to designing and building robust, real-world Erlang applications.

常见问题解答

「Erlang OTP 案例研究」课时是免费的吗？

是的 — 「Erlang OTP 案例研究」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课程的其余内容，请升级到 CoddyKit PRO。 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课程共包含 4 节课。

「Erlang OTP 案例研究」这节课中我会学到什么？

分析大规模 Erlang OTP 部署中的真实案例和最佳实践，从业界成功经验中学习。你通过在浏览器中直接运行的动手代码来练习 Erlang OTP: Distributed & Fault-Tolerant Systems Programming，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 需要有经验吗？

无需任何先前经验。CoddyKit 上的 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 3 节课，共 4 节。

「Erlang OTP 案例研究」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课中编写并运行代码吗？

能。每节 Erlang OTP: Distributed & Fault-Tolerant Systems Programming 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 Erlang OTP: Distributed & Fault-Tolerant Systems Programming