Summary of interview questions (2) IO model, set, NiO principle, cache penetration, breakdown avalanche

1. several IO Model ,

        Blocking IO、 Non blocking IO、 Multiplexing IO、 Signal driven IO, asynchronous IO    （ In my previous article, I explained in detail IO Several models of Blocking , Non blocking ,io Multiplexing , Signal driven and asynchronous io_ It's xiaoqingqing's blog -CSDN Blog

2.Java Collection classes  

 Java The set classes in are divided into 4 Categories: , By 4 Interfaces to represent ,

  They are Set、List、Queue、Map. among ,Set、List、Queue、 Inherit from Collection Interface .

  Set Stands for disordered 、 A collection of elements that are not repeatable .

  List Stands for orderly 、 A collection of elements that can be repeated .

  Queue First in first out （FIFO） Queues .

  Map Represents a mapping relationship （key-value） Set .

 Java Provides many collection implementation classes , They are all direct or indirect implementation classes of these interfaces , One of the more common ones is ：HashSet、TreeSet、ArrayList、LinkedList、ArrayDeque、HashMap、TreeMap etc. .

The interface or implementation of the collection class mentioned above , All in java.util It's a bag , Most of these implementations are non thread safe . Although not thread safe , But the performance of these classes is better . If you need to use thread safe collection classes , You can use Collections Tool class , This tool class provides synchronizedXxx() Method , You can wrap these collection classes into thread safe collection classes .

 java.util In the collection class under the package , There are also a few thread safe collection classes , for example Vector、Hashtable, They are very old API. Although they are thread safe , But the performance is very poor , It's not recommended anymore .

 3.NIO Implementation principle of

      （Buffer、Channel、Selector）

   NIO Is based on IO Implementation of multiplexing model , It contains three core components , Namely Buffer、Channel、Selector.

1. NIO yes Facing the buffer Of , stay NIO All data in the is processed through a buffer .Buffer Is the buffer object , Whether reading or writing , Data is entered first Buffer Of .Buffer The essence of is an array , Usually it's an array of bytes , It can also be other types of arrays .Buffer It's an interface , Its implementation classes are ByteBuffer、CharBuffer、ShortBuffer、IntBuffer、LongBuffer、FloatBuffer、DoubleBuffer.

2. Channel It's a passage , It can be used to read and write data . Unlike flow , The flow is unidirectional , and Channel yes two-way Of . Data can be obtained by Channel Read Buffer in , It can also be done through Channel Write to Buffer in . To support different devices ,Channel Interfaces have several subclasses , Such as FileChannel Used to access disk files 、SocketChannel and ServerSocketChannel be used for TCP Protocol network communication 、DatagramChannel be used for UDP Protocol network communication .

3. Selector It's a multiplexer , It can monitor Listening to the Internet IO The state of . It can continuously poll registered Channel, If a Channel There is a connection on 、 Read 、 The write event occurred , Then this Channel It 's ready , Will be Selector Poll it out . All polled Channel aggregate , We can go through SelectionKey Get , And then we do the following IO operation .

4. Cache penetration 、 Cache breakdown 、 What's the difference between cache avalanches , How to solve ？

  Cache penetration ： The client queries data that doesn't exist at all , Make the request go directly to the storage layer , The load is too large , Even down . The reason for this , It may be that the business layer mistakenly deleted the data in the cache and Library , It could also be a malicious attack , Special access to data that does not exist in the library .

Solution ：

1.  Caching empty objects ： After a storage layer miss , Still store null values in the cache layer , When the client accesses the data again , The cache layer directly returns a null value .

2.  The bloon filter ： Store the data in the bloon filter , Block with filter before accessing cache , If the data does not exist, it directly returns a null value .

Cache breakdown ： A hot data , It's very visited . At the moment of its cache failure , A large number of requests go directly to the storage layer , Cause the service to crash .

·  Solution ：

1.  Never expire ： No expiration time is set for hotspot data , So there won't be any of these problems , This is a “ Physics ” The on the never expires . Or set a logical expiration time for each data , When the data logic is found to be out of date , Use a separate thread to rebuild the cache .

2.  Add mutex lock ： Add mutex to data access , When a thread accesses the data , Other threads can only wait . After this thread access , The data in the cache will be rebuilt , At that time, other threads can take values directly from the cache .

Cache avalanche ： At some point , The cache layer cannot continue to provide services , Cause all requests to go directly to the storage tier , Cause database downtime . It may be that a large amount of data in the cache expires at the same time , May be Redis Node failure , Resulting in a large number of requests that cannot be processed .

·  Solution ：

1.  Avoid data expiration at the same time ： When setting expiration time , Add a random number , Avoid a lot of key At the same time .

2.  Enable degradation and fusing measures ： In the event of an avalanche , If the application is not accessing the core data , The predefined information is returned directly / Null value / error message . Or in the event of an avalanche , For requests to access the cache interface , The client does not send the request to Redis, I'm going straight back .

3.  Build highly available Redis service ： Use sentinel or cluster mode , Deploy multiple Redis example , Individual nodes are down , It can still keep the whole service available .