• Stack is thread-safe.
• Internal use of arrays to store data, when not enough to double.

• The leaf-node linked list structure of B+ numbers is easier than that of B- numbers for library scanning and range retrieval.

LSM (Log-Structured Merge Tree) Storage Engine

• At the extreme, the write performance of LSM tree-based HBase is an order of magnitude higher than that of MySQL, and the read performance is an order of magnitude lower.
• Optimization method: Bloom filter to replace binary search; Compact small digit tree to improve query performance.
• In Hbase, when the memory size reaches a certain threshold, flush is flushed to the disk to form a file (B+ number). HDFS does not support update. Therefore, merge update is performed in Hbase instead of merge update. Flush small trees onto disk, periodically merging into a big tree.

• Bucket sorting divides the [0,1) interval into n subintervals of the same size, called buckets.
• Sort each pass individually, then iterate over each bucket again.

Implementing Binary Lookup in Java – Two Ways

• While + recursion.

• Bitmap data structure based on Redis.

• Use the Java BitSet class and the weighted sum hash algorithm.

Bayesian Inference and Its Internet Application 1

Bayesian Inference and Its Internet Application ii

Four features of database transactions and transaction isolation levels

MySQL InnoDB magic Reading Problem

• The example of illusory reading is very clear.
• Through the SELECT… FOR UPDATE.

An article about MySQL and InnoDB

• Illustrates dirty reading, unrepeatable reading and phantom reading problems.

MYSQL MVCC

• MVCC control is achieved by hiding the version columns, one for creation time, one for deletion time, and here the time
• Only rows smaller than (or equal to) the current version are operated on at a time.

An Implementation of Optimistic Locking — CAS

• This is similar to MySQL optimistic locking, except that it is compared to the original value.

• AtomicStampedReference and AtomicStampedReference.

Copy-on-write Containers in Java

ReenTrantLock ReenTrantLock (the difference between synchronized)

• Synchronized is easy to use, the compiler to lock, unfair lock.
• ReenTrantLock is flexible and the fairness of the lock can be customized.
• Synchronized is recommended in the same locking scenario.

How does Java view deadlocks?

• JConsole recognizes deadlocks.

Java Multithreading series: Deadlocks and detection

• Jstack can show deadlocks.

Spring- Summary of Design Patterns Involved

Design Patterns used by Mybatis

• Use enumerated types.

• Forward control: traditional via New. Reverse control, injecting objects through the container.
• Purpose: Used for module decoupling.
• DI: Dependency Injection, which only cares about resource usage, not source.

• Spring AOP uses dynamic proxies in two main ways: JDK dynamic proxies and CGLIB dynamic proxies.

• The Spring AOP framework treats AOP proxy classes as follows: If the target object’s implementation class implements an interface, Spring AOP will use JDK dynamic proxies to generate AOP proxy classes; If the target object’s implementation class does not implement an interface, Spring AOP will use CGLIB to generate the AOP proxy class

Lecture 20 on The Core of Microservices Architecture

Selection and Practice of Open Source APM Technology

• Mainly based on Google’s Dapper (tracking system for large-scale distributed systems) idea.
• Open source software: Pinpoint, SkyWalking, Zipkin, CAT

Common statistical tools, targets and contents of buried points in APP

• Third-party statistics: Umeng, Baidu Mobile, Rubik’s Cube, App Annie, Talking Data, Shence Data, etc.

Meituan-dianping Front-end Traceless Embedding Practice

• The so-called traceless means that the nodes are configured and collected by visual tools, and the configuration is automatically analyzed and the buried point data is reported in the front end, instead of hard coding.

• TestNG overrides JUnit and is suitable for more complex scenarios.

• Module interface test, local data structure test, path test, error handling test, boundary condition test.

Nginx vs. Apache: Pros and Cons

• Nginx is only suitable for static and reverse proxies, not for dynamic requests.

• Development can be done on Nginx through the Lua module.

Tomcat HTTP protocol and AJP Protocol

AJP vs. HTTP Comparison and Analysis

• The AJP protocol (port 8009) is used to reduce the number of connections to front-end servers (such as Apache) and to improve performance through long connections.
• When the concurrency is high, AJP is better than HTTP.

• Architecture comparison :Jetty’s architecture is simpler than Tomcat’s.
• Performance comparison: There is little difference between Jetty and Tomcat in terms of performance. Jetty uses NIO by default to process I/O requests. Tomcat uses BIO by default to process I/O requests.
• Other aspects: Jetty is faster to use, easy to modify, and has good support for the new Servlet specification; Tomcat has more comprehensive JEE and Servlet support.

EhCache Local Cache

• In-heap, off-heap, disk level cache.
• The value can be set based on the cache capacity.
• Expiration policy based on time and number of times.

The Guava Cache”

• Simple and lightweight, no out-of-heap, disk cache.

The Guava Cache”

• Simple and lightweight, no out-of-heap, disk cache.

Nginx Local Cache

Pagespeed — Lazy People’s Tools, Server-side Acceleration

Analysis and Practice of WebView Caching Mechanism on H5 and Mobile Terminal

Understanding Memcached in depth

• Use multiplexing to improve concurrency.
• Slab allocation algorithm: memcached allocates memory space to a slab. The default value is 1MB. After a Slab is allocated, the Slab is divided into chunks of the same size. Chunk is the memory space for caching records, and the size of chunk increases by 1.25 times by default. The advantage is that it does not frequently apply for memory, which improves I/O efficiency. The disadvantage is that there will be a certain amount of memory waste.

How Memcached Software Works

Memcache Technology Sharing: Introduction, Use, Storage, Algorithm, Optimization, Hit Ratio

The difference between Add, Set and Replace in memcache

• The difference is that the return value is true and false when the key is present or absent.

Full Anatomy of Memcached

Redis Fundamentals

• Use Ziplist to store linked lists. Ziplist is a kind of compressed linked list. The advantage of ziplist is that it can save more memory space, because it stores the contents in the contiguous memory area.
• Use Skiplist (skip table) to store ordered collection objects, search from high Level first, time complexity as red black tree, easy to implement, no lock, good concurrency.

Redis Persistence

• RDB: Periodically back up snapshots for disaster recovery. Advantages: Backup by fork does not affect the main process. RDB can restore large data sets faster than AOF. Disadvantages: Loss of data.
• AOF mode: saves operation logs. Advantages: Less data is lost during recovery. Disadvantages: Large files and slow recovery.
• You can also use a combination of both.

Distributed Caching — Sequence 3– Atomic Operations and CAS Optimistic Locking

“Comparison of Kafka, RabbitMQ, RocketMQ and other message middleware — Message sending Performance and Differences”

How To Run Linux Cron

• Fork process + sleep polling

Quartz Usage Summary

—- Trigger on time startup principle

The Fundamentals of Quartz and the source code

• In the QuartzSchedulerThread code, while() loops indefinitely, each loop takes out the trigger whose time is about to expire and triggers the corresponding job until the scheduler thread is closed.

Basic Implementation Principles of Quartz Task Scheduling

• In a Quartz cluster, independent Quartz nodes do not communicate with another node or management node, but are aware of another Quartz application through the same database table

“Performance Comparison of Distributed RPC Framework with Dubbo, Motan, RPCX, gRPC and Thrift”

• Supports multiple languages and defines interfaces through intermediate languages.

Unified Configuration Management Based on ZooKeeper

Spring Cloud Config Distributed Configuration Center Tutorial

On the Background, Architecture and Implementation Scheme of API Gateway

“Building API Gateways with Zuul”

HTTP API Gateway: An Introduction to Kong

• Use binary frames for transmission.
• Multiplexing.

A summary of the differences between Select, Poll and epoll

• Select, poll, and epoll are essentially synchronous I/O because they are responsible for the read and write after the read and write event is ready, that is, the read and write process is blocked.
• Select open file descriptor limit, default 1024 (2048 for x64), 1 million concurrent, will use 1000 processes, switching overhead; Poll uses a linked list structure and has no number limit.
• Select, poll traverses the entire FD collection when it is “awake”, while ePoll only needs to determine whether the ready list is empty when it is “awake”, saving a lot of CPU time through the callback mechanism. Select, poll copies the FD collection from user state to kernel state once per call, whereas epoll only copies it once.
• With the increase of concurrency, poll performance gradually decreases. Epoll adopts a red-black tree structure and its performance is stable and does not decrease with the increase of connection number.

Select, Poll, epoll Comparison

• Select and Poll may perform better than epoll when the number of connections is small and the connections are very active, because epoll’s notification mechanism requires many function callbacks.

Understanding Java NIO in Depth

• NIO is a synchronous non-blocking IO model. Synchronous means that the thread continuously polls for I/O readiness. Non-blocking means that the thread can do other tasks while waiting for the I/O

BIO: The Difference between NIO and AIO

Two Efficient server design models: Reactor and Proactor models

The best 20+ Tips for MYSQL Performance Optimization

The Way of SQL Optimization

Mysql Database deadlock causes and Solutions

“Possible Conditions leading to Index failure”

MYSQL pagination limit is too slow

• The principle is to narrow down the scan.

In-depth Learning HBase Architecture Principles

Differences between Traditional Row Storage and (HBase) Column Storage

Differences between Hbase and Traditional Databases

• Empty data is not stored, saves space, and is suitable for concurrency.

HBase Rowkey Design

• Rowkeys are arranged in dictionary order to facilitate batch scanning.
• Hot spots can be avoided by hashing.

Detailed Explanation of the Implementation principle of Sliding Verification Code

• Slide captcha evaluates risk based on the person’s response time on the slide, drag speed, time, position, trajectory, number of retries, etc.

Research on Taobao Sliding Verification Code

Public URL Signature Verification Design Based on Hash Digest Signature

About Highly Available Systems

• Design principle: no data loss (persistence); High availability of services (service replicas); Absolutely 100% high availability is hard, the goal is to do as much 9 as possible, like 99.999% (only 5 minutes accumulated throughout the year).

What do you Know about Software/Hardware Load balancing Products?

Nginx Load Balancing

• Simple and lightweight, low learning cost; It is mainly applicable to Web applications.

Load Balancing with LVS+Keepalived

• Configuration load comparison, support only to layer 4, high performance.

The most detailed Chinese document of HAProxy usage

• Support to layer 7 (such as HTTP), more comprehensive features, performance is good.

Haproxy+Keepalived+MySQL to achieve read load Balancing

• It is mainly load balancing of user read requests.

Rabbitmq + HaProxy + Keepalived high Availability cluster building

Cache Penetration, Cache Breakdown, Cache Avalanche Solution Analysis

Cache breakdown, failure, and Hot Key Issues

• Main strategy: failure moment: single use lock; Use distributed locks; Not expired;
• Hotspot data: The hotspot data is stored separately. Use local caching; Divided into multiple subkeys;

Living in Different Places (Living in different Places) practical Experience

• Be aware of delays, which can be magnified several times by multiple cross-room calls.
• Building room special line is very likely to have problems, do a good job of fault tolerance operation and procedure level.
• Do not rely on program – side data double – write, have automatic synchronization scheme.
• Data is never considered synchronization quality under high latency and poor network quality.
• Divide and conquer the core business and the secondary business, even consider the core business.
• Remote live monitoring deployment, testing should also keep up.
• Consider user segments if business allows, especially gaming and email services.
• Control the size of message body across machine room, the smaller the better.
• Consider using Docker container virtualization technology to improve dynamic scheduling capability.

Dr Technology and construction experience

MySQL Master-slave Replication

Load Balancing by Haproxy+ Multiple MySQL Slave Servers

DRBD+Heartbeat+Mysql High Availability Read-write separation Architecture

• DRDB performs disk replication to avoid single point problems.

MySQL Cluster Mode

MySql > alter table partition;

• Partition: an internal mechanism of MySQL that is transparent to clients. Data is stored in different files, ostensibly in the same table.
• Split table: Different tables are created physically, and the client needs to manage the split table routes.

Distributed Lock Based on Zookeeper

• Clear principle description + Java code examples.

JedisLock — Redis Distributed Lock Implementation

• Based on setnx(set if ont exists), return false if it exists and true otherwise. Expiration time is supported.

Memcached and Redis Distributed Locking Scheme

• Using memcached’s add (as distinct from set) operation, return false if the key exists.

Comparison of advantages and Disadvantages between DDD CQRS Architecture and traditional Architecture

• Ultimately consistent design philosophy; Rely on high availability messaging middleware.

Introduction to CQRS Architecture

• An abstract example of implementing CQRS.

Long in Depth: My Thoughts on CQRS/EventSourcing Architecture

• CQRS mode analysis + 12306 ticket snatching cases

What Architects Need to Know about Architecture Optimization and Design

• Details to consider: modular, light coupling, shared-nothing architecture; Reduce the lag before each component, and pay attention to the chain of failures and impacts caused by lag between services.
• Comprehensive consideration of infrastructure, configuration, testing, development, operation and maintenance.
• Consider the impact of people, teams, and organizations.

How can I Really Improve myself and become an Excellent Architect?

“Essential Qualities and Growth Path for Architects”

• Quality: business understanding, technical breadth, technical depth, rich experience, communication skills, hands-on skills, aesthetic literacy.
• Growth path: 2 years to accumulate knowledge, 4 years to accumulate skills and ancestral influence, 7 years to accumulate influence within the department, more than 7 years to accumulate cross-department influence.

Architect – What Floor Are You on?

• The tier 1 architect sees only the product itself
• The second-tier architect sees not only his product, but the overall solution
• Tier 3 architects see business value

ITeye

• Java by

Blog garden

ChinaUnix

• Partial Linux direction

Open Source Chinese Community

The depth of the open source

Bole online

• Covers IT workplace, Web front end, back end, mobile end, database and other aspects of the content, technical end.

ITPUB

Tencent Cloud — Cloud + Community

Ali Cloud — Cloud Habitat Community

IBM DeveloperWorks

Developer headlines

LinkedKeeper

Operations:

• Ops sent
• Abcdocker

Java:

• ImportNew
  • Focus on Java technology sharing
• HowToDoInJava
  • English blog

security

• Red and black alliance
• FreeBuf

Big data

• Big Data in China

Other thematic websites:

• DockerInfo
  • Focus on Docker application and consultation, tutorial website.
• The Linux community
  • Linux Themed Community

Runoob.com

• Introduction to HTML, CSS, XML, Java, Python, PHP, design patterns, etc.

Love2.io

• Many, many Chinese online e-books, is a new open source technical document sharing platform.

gitbook.cn

• Paid e-books.

ApacheCN

• AI, big data series of Chinese documents.