Shulou(Shulou.com)06/01 Report--

Our goal is

Understand the Oracle database management system and comb through the core keywords in series.

Database management system is like a network, the key word is the intersection of the Internet. We will not explain in detail the meaning of each keyword, Baidu or official documents will have a detailed description, so far, there is a full picture, a familiar face.

Those who don't know, have an impression, those who know, smile clearly. That's all.

Let's talk about Oracle first, but before we talk about Oracle, we have to mention IBM.

IBM (International Business machines Corporation) or Universal Business machines Corporation, IBM (International Business Machines Corporation)

The name is so majestic, so far it is like thunder.

The key is that there are a large number of awesome people in it.

First of all, this one:

His name is E.F.CODDE. By the way, his name is a very important key word. Because he is the Dharma ancestor of the relational database.

In 1970, E.F.CODDE published "Relational Model of large-scale shared Database data", which clearly and clearly put forward a new model for database system for the first time, namely relational model.

There was a clap of thunder and a rock-shattering sound.

Followed by Tang, who is also a great man of IBM. Qian Berlin (Don Chamberlin) took the stage. He is the father of SQL. E.F.CODD only solves the study of relational algebra at the mathematical level, while Don Chamberlin translates relational algebra into a language that can be easily implemented in computers, that is, SQL.

The so-called relational algebra includes

Set operations such as union, difference, intersection, Cartesian product, etc.

And

Selection, projection, join, natural join, outer join, division and other relational operations.

The SQL language is all too familiar to us. Without it, a relational database is a backwater.

SQL is as important as a relational database!

The theory is available, and then there is the question of how to use it. Of course, issues such as SQL standardization are a matter of gradual competition and take time to precipitate. All computer protocols and standards can be viewed in this way. Finally, books and articles, cars on the same track is the trend of the times, but also inevitable, otherwise, disorderly, can only be scattered, talk about the Internet of everything.

A man with a keen sense of business is moved by the wind.

He is Larry Ellison (Larry Ellison) He formed a team of four, as shown below:

The one on the far right is Larry Ellison, and the one holding the cake is Bruce Scott.

A reserved test account in the Scotte,Oracle database, password tigger, is the name of his cat.

It is difficult to start a business, but it can also be said to be smooth.

Oracle caters to the trend and stands at the top of the tide.

The picture below is the headquarters of Oracle in California today. It can be seen that the momentum is magnificent and the wealth is deep. Although Oracle has gone through mergers and acquisitions and multi-dimensional development, it is not only a database manufacturer, but the database-shaped building also indicates that it started with the database.

Besides, the database does not need too much explanation, as the name implies, the warehouse that stores the data. But with the development of the Internet, today, the database is divided into relational database and non-relational database, Oracle database, that is, relational database, of course, including Mysql,MS-SQL Server,PostgreDB. Non-relational database, it is a bit mixed, a hundred schools of thought contend with the characteristics. Such as MangoDB,HBASE,Redis, Cassandra. Although explaining them, boasting that NoSQL is a new database revolutionary movement, what "Not only SQL", I think that non-relational databases are not so great and epoch-making compared to relational databases. It is just a change and improvement to the relational database. Because the relational database is rigorous and thorough, it will carry a heavy load. For some application scenarios, you don't need to be so rigorous, you just need to travel light and achieve the required functions. What ACID, data consistency, paradigm and so on will become cumbersome. So non-relational databases, to put it simply, have the following characteristics:

1. For specific application scenarios, the function is simplified.

two。 The ACID mechanism is not supported, which is also a cornerstone of relational databases.

3. It is relatively easy to use, and there is no architecture as large as Mysql,Oracle.

4. Most of them are open source.

I remember more than a decade ago, MySQL is still a small and medium-sized database, simple, does not support transactions, read and write or blocking, and so on, as a MySQL DBA is very low, and as follows, Mysql DBA recruitment information is dazzling, and well-paid, Yanran has overwhelmed Oracle DBA. Because now Mysql,InnoDB is owned by Oracle, the two are closely combined, and the energy level is doubled. Relational database products tend to be the same. Moreover, many problems have become an architectural problem, and many deficiencies in Mysql can be solved through architecture. Of course, free is also a problem, but for many companies with less complex businesses, they can't use hundreds of database servers (like Amazon, the cost of buying database products is not to be underestimated). The cost of database products is a drop in the bucket for them. And stable and efficient service support is what they should pay more attention to, and stability trumps everything. So Oracle is still their first choice.

Therefore, with the growth and improvement of open source relational database products, and the contention of a hundred schools of thought in NoSQL database, Oracle can no longer be said to have lost its glory, but it can no longer be said to be in its heyday.

As a DBA, you should know more.

Like treading on thin ice, trembling!

Worried and conscientious!

A sense of suffering is necessary for every DBA, especially with the rapid development of technology today.

Besides, database management system.

The official document says:

A database management system (DBMS) is software that controls the storage, organization, and retrieval of data.

Stroage (storage), Organization (organization), Retrieval (retrieval)

First of all, it is a set of software, but it is huge, complex, and a system-level software. But after all, it is still a software, with the general characteristics of the software.

Software is a collection of a series of programs. The program is dead, it is a static file, and it needs to be made to work. When a program is loaded into memory and becomes active in memory, it behaves as a process. Therefore, it depends on CPU, memory and other resources. The process is not a vegetarian meal, and then die for nothing. Computers do not support idle people, take up space, do not work or do not do business, that is the virus. Useful processes are to work, to blossom and bear fruit, and most of the time these results need to be stored and persisted. This is what external memory is for. After all, it is a database, not only data, but also a library, how can we float without a trace and not do persistent storage.

Therefore, I should pay attention to:

Memory, CPU, storage, process state, which are factors at the operating system and hardware level, require constant attention in day-to-day monitoring and optimization.

Of course, when building an environment (including hardware and software), we should pay more attention to these factors. In particular, the storage architecture, Iamp O throughput, must be guaranteed, and Imax O shunt balance is also very important. After all, the database is an Imax O-intensive application. A good beginning is half the success, with a good structure, a hopeful back arch and a poor structure, and it is hard to cope with it.

"wipe frequently so as not to stir up dust."

Post a picture of "Brother Bird's Linux private dish". It's basic, but it can't be ignored.

When we get inside the database, we should focus on:

Memory structure, process state and function, and storage structure

These are the basic routines of learning any set of software, because the underlying layer of the computer is the same, memory, CPU,I/O, external storage.

The following pictures are basically keywords that you can savor carefully. For beginners, if you want to understand the principle, it is best to look at the picture and learn through some videos, learn it quickly, through official documents or second-hand instructions, maybe in a fog.

Memory structure and processes:

Storage structure (physical structure): mainly a series of files:

Key files: control files, data files, and online redo log files.

The control file is the hub, the data file is the basic, and the online redo log file is the guarantee

One of the three major documents is indispensable.

Non-critical files: parameter files, password files, archive log files, alarm files, tracking files, backup files

Storage structure (logical storage):

What needs to be noted here is the data block.

This is what the official said:

A data block is the smallest logical unit of data storage in Oracle Database.

One logical data block corresponds to a specific number of bytes of physical disk space, for example, 2 KB. Data blocks are the smallest units of storage that Oracle Database can use or allocate.

A block is the smallest logical unit in which data is stored in an Oracle database.

A logical data block corresponds to a specific number of bytes of physical disk space, for example, 2KB. A block is the smallest unit of storage that an Oracle database can use or allocate.

It's like a box with our stored data, that is, a row of data in a table. Every time you extract data from disk, you pick up the box and put it in memory. The corresponding in memory is buffer, which is a box in memory. What is stored in the computer is 0 and 1, so the box, or block, is just a logical concept of encapsulation. Block is a logical concept, but it is the basis of all concepts in storage.

Istroke O throughput is in blocks.

Performance metrics are essentially a tradeoff between how many blocks you get and how much time you spend.

Encapsulation because it needs to be transmitted. It's like mailing a package to be packed. Address information should be written on the package, which is all too familiar to us in TCP/IP. It is packed and packaged layer by layer. Free delivery package head, block has a big head.

Size is very important. Because it contains transaction information, lock information, etc., which is the basis of relational data.

Everything has a head, a block has a head, a block has a head, a section has a head, this kind of routine, in the computer, has been around for a long time, even born, so it can not be ignored.

Finally, when it comes to databases, we are talking about relational databases. Then we must know what a relationship is.

The relational model proposed by E.F.CODD. We can't understand the mathematician's paper, but someone can read it for us.

A relationship is a two-dimensional table.

Scientists have done a series of work in defining relationships, and there is no need for us to follow them. We can certainly try if we are interested.

Scientists have designed a series of specifications to define relationships and to solve the problems of redundancy and consistency in database design.

The so-called norms can be obeyed, because scientists have demonstrated that this is right.

Specification is the paradigm, in most cases, we design the database should follow the specification, in line with the requirements of the paradigm. But only in most cases, in some cases, in order to cater to some special cases, it is necessary to do anti-paradigm design.

The so-called anti-paradigm design means deliberate redundancy, handing the consistency problem to the application and business logic. Trade space for time.

When it comes to data consistency, it can be said that structured data is a cornerstone of relational databases. A person to modify a data, there is no inconsistency, but the database management system can not be a single operation, or how can it be said that it is huge and complicated? It is inevitable for many people to visit the revision together, that is, to send it at the same time.

Concurrency is the natural enemy of data consistency. The truth is self-evident.

It is not easy for the two to live in harmony. So the lock made its debut. It is a regulator of concurrency and data consistency.

But locks are made for adjustment. Adjustment is actually very rough.

Philosophically speaking, the absolute of time and the relative of space. At the same time, a data structure can only be modified by one person. Lock, that is, let others wait, waiting will cause blockage. So the performance problem arises.

If there is no lock, there is no wait (of course, some idle waiting is not in this consideration), there is no blocking, and the database will run smoothly, but this is not possible in a relational database system.

However, in non-relational databases, this is possible because they may give up locks, so why can they? Because they eliminate disputes over the same resource, that is, do not concurrency. Because the business doesn't need it. In line with business needs, but also able to walk as fast as flying.

End

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.