從 Java 中的零拷貝到五種IO模型

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在之前的文章中，我們聊過了 Java 中的零拷貝，零拷貝就是指數據不會在內核空間和用戶空間之間相互拷貝。這樣就減少了內核態與用戶態的切換，自然就很高效。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"拷貝文件只是 IO 操作中一個特殊的情況，大多數的 IO 操作還是需要將數據從內核空間移到用戶空間，這往往是一個比較耗時的操作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"IO 操作不僅僅指對文件的讀寫，網絡的通信同樣也是 IO 操作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如今很多系統的瓶頸就在於 IO 上，比如經典的 C10K，C10M 問題本質上就是在解決 IO 問題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這篇文章將介紹經典的 IO 模型的實現原理，以及說明 Java IO 與這幾種 IO 模型的關係。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"💡本文討論的環境爲 Linux"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"IO 操作是怎麼實現的"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"IO 操作是一個很複雜的過程，遠遠不止調用一個函數那麼簡單，因爲每一次的 IO 操作都會涉及到操作系統的內核空間和用戶空間的轉換，真正執行的 IO 操作實際上是在操作系統的內核空間進行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這是一個很耗資源的操作。計算機中內存和 CPU 都是非常稀有的資源，應該儘可能提高這些資源的使用效率。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"IO 操作經常需要與磁盤就行交互，所以IO 操作相比於 CPU 的速度要慢好幾個數量級。利用這兩者之間的速度差異，就可以實現不同種類的 IO 方式，也就是俗稱的 IO 模型。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，這些 IO 操作的都在操作系統層面上實現好了，編程語言可以利用這些能力去實現 IO 相關的 API。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"五種 IO 模型"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 操作系統中，IO 模型有如下五種："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/97/97182b7b0e5ee15aaf84dc82a9a26812.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"03-08-93-%E4%BB%8E%20Java%20%E4%B8%AD%E7%9A%84%E9%9B%B6%E6%8B%B7%E8%B4%9D%E5%88%B0%E4%BA%94%E7%A7%8DIO%E6%A8%A1%E5%9E%8B%20aadacfd890fd46e6ad698ea27c9cffae/Untitled.png","title":null},"marks":[{"type":"underline"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在上文已經說到，其實 IO 操作就是將數據在用戶空間與內核空間進行相互轉換，這個過程是通過`系統調用`來完成的。 IO 技術的發展目標就是如何使用盡可能少的資源來完成數據的傳輸，這裏資源主要就是指 CPU 資源。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"無論是文件 IO，還是網絡 IO，最後都可以統一爲用戶空間和內核空間數據的交換。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"BIO"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"BIO 是最經典的一種 IO 方式，也是最簡單粗暴的方式，在發起 IO 操作之後，當前調用線程就會處在阻塞狀態，直到數據傳輸完成。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/a7/a792571b3500e39de14a0a415fb6301a.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"NIO"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"NIO 是在 BIO 基礎之上的一個改進，NIO 在數據還未準備好的情況下，不會阻塞進程，而是通過輪詢的方式，不斷的去查詢數據時候準備好，當數據可以被讀取時，當前線程就會處在阻塞狀態，直到數據讀取完成。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/bf/bf5f0a1c7aef722436940a119d2a3b35.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以 NIO 中的非阻塞指的是在等待數據的階段，實際進行數據傳輸時，還是阻塞的，這點需要注意。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"IO 多路複用"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"IO 多路複用是對 NIO 的一個改進，在 NIO 中，需要不斷輪詢查看數據是否準備好，IO 多路複用的改進是不再主動去查詢數據狀態是否準備完成，而是等數據準備好的通知，當數據準備完成之後，纔會開始傳輸數據。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/9d/9dbce77bc87c81d68abeb75743d3bab6.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"與 NIO 一樣，在數據的傳輸階段，當前線程依然是阻塞的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Linux 系統中，IO 多路複用的方式有多種："}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"select"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"poll"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"epoll"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"信號驅動 IO"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"信號驅動 IO 通過 sigaction 系統調用，向內核發送一個信號，當內核中數據準備好之後，當前線程也會接收到一個信號，在這個過程中，當前線程也是非阻塞的。在接收到信號之後，就可以開始傳輸數據。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/18/18ced1776ad0db587bd0dc83676d6d27.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"AIO"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"上面的這些 IO 模型雖然有些號稱是不阻塞的，那是指在等待數據就緒的過程中是不阻塞的，但是在接收數據的時候，依然還是阻塞的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AIO 是這些 IO 模型中真正實現完全不阻塞，AIO 在被調用之後直接返回，連接收數據的階段也是非阻塞的，等到數據接收完成之後，內核纔會返回一個通知，也就是說當用戶進程接收到通知時，數據已經接收完成。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/80/80017d16e9749b90f97beffb404cdebc.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Linux 中提供了 AIO 的實現，但是實際上使用的並不多，更多還是使用獨立的異步 IO 庫，比如libevent、libev、libuv。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"五種 IO 模式的總結如下："}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/55/55b2d6d82ff910584915d453fc706123.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Java 中 IO 的實現"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Java 中的 IO 也不例外，實際的 IO 是調用了系統的能力來完成，在用戶態通過系統調轉到內核態，最終實現文件的讀寫或者通信。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Java 中 IO 就是典型的 BIO，而且 NIO 則不是對應五種 IO 模型中的 NIO，Java 中的 NIO 實際上是使用 IO 多路複用來實現的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Java 中的 NIO2 也稱之爲 AIO，正是對應操作系統中的 AIO，當然具體的實現可能是其他的庫。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#F5222D","name":"red"}}],"text":"文 / Rayjun"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"REF"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"[1] Unix 網絡編程"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本文首發於我的微信公衆號"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":""}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/af/aff0781fde8a41d0b23e298cc79993c7.jpeg?x-oss-process=image/resize,p_80/auto-orient,1","alt":null,"title":"歡迎關注我的微信公衆號","style":[{"key":"width","value":"25%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}