Myisam is preferred without transaction and little update(delete)

Big than 4G datafile can user Myisam merge table.

InnoDB with auto_increment primary key is preferred.

Few storage process

Guess: 20m records max per table , 500G data max per tablespace , 256 tables per database (may problem)

Use prepared statement and  batch

Optimize Your Queries For the Query Cache

// query cache does NOT work

$r = mysql_query("SELECT username FROM user WHERE signup_date >= CURDATE()");

// query cache works!

$today = date("Y-m-d");

$r = mysql_query("SELECT username FROM user WHERE signup_date >= '$today'");

EXPLAIN Your SELECT Queries

LIMIT 1 When Getting a Unique Row

Index and Use Same Column Types for Joins

Do Not ORDER BY RAND()

Avoid SELECT *

t is a good habit to always specify which columns you need when you are doing your SELECT’s.

Use ENUM over VARCHAR

Use NOT NULL If You Can

Store IP Addresses as UNSIGNED INT (?)

Fixed-length (Static) Tables are Faster

Vertical Partitioning

Vertical Partitioning is the act of splitting your table structure in a vertical manner for optimization reasons.

Example 1: You might have a users table that contains home addresses, that do not get read often. You can choose to split your table and store the address info on a separate table. This way your main users table will shrink in size. As you know, smaller tables perform faster.

Example 2: You have a “last_login” field in your table. It updates every time a user logs in to the website. But every update on a table causes the query cache for that table to be flushed. You can put that field into another table to keep updates to your users table to a minimum.

But you also need to make sure you don’t constantly need to join these 2 tables after the partitioning or you might actually suffer performance decline.

Split the Big DELETE or INSERT Queries

If you have some kind of maintenance script that needs to delete large numbers of rows, just use the LIMIT clause to do it in smaller batches to avoid this congestion.

Smaller Columns Are Faster

Use an Object Relational Mapper

f you do not need the time component, use DATE instead of DATETIME.

Consider horizontally spitting many-columned tables if they contain a lot of NULLs or rarely used columns.

Be an SQL programmer who thinks in sets, not procedural programming paradigms

InnoDB can’t optimize SELECT COUNT(*) queries. Use counter tables! That’s how to scale InnoDB.

Prefer MM with hive

refer :

http://blog.tuvinh.com/top-20-mysql-best-practices/

PreparedStatements are implemented by the driver, as MySQL does not have a prepared statement feature. Because of this, the driver does not implement getParameterMetaData() or getMetaData() as it would require the driver to have a complete SQL parser in the client.

Starting with version 3.1.0 MySQL Connector/J, server-side prepared statements and binary-encoded result sets are used when the server supports them.

声明,因为c++这种语言,在使用函数和变量的时候,必须将他声明,为何要声明?声明之后才

知道他的规格,才能更好的发现不和规格的部分.你别妄想一个编译单元,会自动从另一个

编译单元那里得到什么信息,知道你是如何定义这个函数的.

    所以说,只要使用到该函数的单元,就必须写一份声明在那个.cpp里面,这样是不是很麻烦,

而且,如果要修改,就必须一个一个修改.这真让人受不了.

.h就是为了解决这个问题而诞生,他包含了这些公共的东西.然后所有需要使用该函数的.cpp,只需要

用#include包含进去便可.以后需要修改,也只是修改一份内容.

请注意不要滥用.h,.h里面不要写代码,.h不是.cpp的仓库,什么都塞到里面.

如果在里面写代码,当其他.cpp包含他的时候,就会出现重复定义的情况,

比如将函数func(){printf}；放到头文件a.h,里面还有一些a.cpp需要的声明等；

然后你发现b.cpp需要用到a.cpp里面的一个函数,就很高兴的将a.h包含进来.

注意,#include并不是什么申请指令,他就是将指定的文件的内容,原封不动的拷贝

进来.

这时候实际上a.cpp和b.cpp都有一个func()函数的定义.

如果这个函数是内部链接static的话,还好,浪费了一倍空间；

如果是extern,外部链接(这个是默认情况),那么根据在同一个程序内不可出现

同名函数的要求,连接器会毫不留情给你一个连接错误!

http://www.cnblogs.com/shelvenn/archive/2008/02/02/1062446.html

Platform
Kernel
Framework

Be aware of context
Extreme maintenance
Be pragmatic
Extreme abstract: Program to an interface (abstraction), not an implementation
  
Extreme separation of concerns
Extreme readability
Testability
No side effect
Do not repeat yourself

DIP ，dependency inversion of control
OCP , open close
LSP , liskov substitute
ISP , interface segregation
SRP , single responsibility
LKP, Lease knowledge principle

Construction
Behavior
Structure

Long method
Diverse change
    Repeated code
    Talk to stranger
    Pre optimize

 nLongN
 Divided and conqueror

Hierarchal
Pipes and filter
Micro kernel
Broker
Black Board
    Interpreter

What to concurrent

Scalability
    Stretch key dimensions to see what breaks

Ruby
Erlang
assemble
C
C++
Java
Python
Scala

Be ware of different program paradigms.

 Minimize remote calls and other I/O
 Speed-up data conversion
 release resource as soon as possible 

软件设计思想的发展逻辑，大致是提高抽象程度 ，separation of concern 程度。
    fn(design )=  fn1(abstraction )+ fn2(separation of concern).

由于大规模数据处理时代的来临，下一代设计范式的重点：
1.    将是如何提高distributed(--concurrent) programing 的抽象程度 和 separation of concern 程度。
2.    dsl ，按照以上的公式，也确实是一个好的方向。

<art agile software development>
<prerefactor>
<design patterns>
<beautiful architecture>
<refactor>
<pattern oriented software architecture>
<extreme software development>
<beautiful code>
<patterns for parallel programming>
<java concurrent programming in practice>
<java performance tuning>
<the definite guide to hadoop>
<greenplum>
<DryadLINQ>
<software architecture in practice>
<97 things architecture should known>
http://en.wikipedia.org/wiki/Programming_paradigm

1. 将是如何提高concurrent programing 的抽象程度 和 seperation of concern 程度。
2. 至于dsl ，我研究不多，不过，按照以上的公式，也确实是一个好的方向。