This article is the author in the process of learning the three levels of database technology, summarizes some of the knowledge, is copied, help you to learn,

There are three kinds of OLAP is implemented: ROLAP and MOLAP, HOLAP,
OLAP data analysis OLTP data processing
The semantic definition of UML modeling in a four layer in the conceptual framework: yuanyuan model layer, the meta model layer, model layer and the user model layer,
Gathered in UML, is a special kind of correlation, the relationship between the class he said the relationship between the whole and part,
In UML, use case diagram is used to describe the interaction between system and external system and user,
In UML, state diagram, state transfer is driven by events, between
In UML, describing the physical configuration of hardware and software in the system and the figure called a deployment diagram of system architecture,
Data throughput refers to the system can be finished in a unit time the number of database transactions unit TPS.
Every TPS value generation is used to measure system performance indicators,
Parallel database, hierarchy: the top (no share structure), the underlying (Shared memory or Shared disk structure),
No share structure is considered to be the best ice careful support parallel database system structure, is suitable for the application of bank teller,
The dividing method of one dimensional data in parallel database as follows:
Rotary method: the most suitable for scanning the entire table;
Hash partitioning method: than rotation method is more suitable for some queries, also suitable for sequential scan relations;
Range partitioning method: obviously conducive to range query and query,
Google's cloud database is a structured data distributed storage system, called Bigtable,
Shard transparency, location transparency, local image transparency,
The log file format, there are mainly to record the log files for the unit and to block as the unit of log files,
The grammar of the BACKUP LOG: BACKUP the LOG database_name (database) TO backup_device (BACKUP),
Transaction log backups are available only in full recovery mode and the large capacity log recovery mode,
Three types of transaction log backups: pure log backups (only contains a certain interval of transaction logging and not included in the large capacity of any large capacity change log restore perform backup); Large capacity log backups (include logging, and performed under large capacity log recovery any backup, large capacity change cannot be timing recovery). End back up (to may have damaged database backups, to capture not backup logging),
DFD the data flow diagram, which is by the data flow, processing, data storage and the external,
A window function formula: the function name (column) OVER (option),
Window function OVER (PARTITION BY T1. The course number)
DECLARE cursor syntax: DECLARE cursor name cursor FOR SQL statement
Extract the cursor line 5 data statement is: FETCH 5 FROM ABSOLUTE cur1
Extract the cursor is equivalent to the current line in 5 rows of data to the first statement is:
5 the FROM the FETCH RELATIVE cur1
Modify the definition of a FUNCTION: ALTER the FUNCTION
To delete a FUNCTION: the DROP FUNCTION
Define the only non clustered index: CREATE UNIQUE NONCLUSTERE
Distributed database, the hybrid subdivision is horizontal fragmentation, vertical fragmentation and export shard mix,
Local database management system, distributed database system consists of a global database management system, global data dictionary, communication management of four parts,
Distributed database is the basic characteristics of three: the centralized management, high availability, local autonomy,
Recovery control USES the distributed database system is the most typical strategy based on two-phase commit protocol,
Knowledge discovery is mainly composed of three steps, they are data preparation, data mining, explain the results of evaluation,
Three main stages: data mining, data preparation, data mining, the result of interpretation and evaluation,
Method called data mining clustering algorithm: for grouping data from a data set, allows data in each group as far as possible is similar and different between different groups of data as possible,
Data mining method, the association rule mining is used to find the correlation between the data in the database,
The establishment of the association rules or not support degree and confidence is commonly used in two indicators,
The relationship between the use cases including extensions, use and combination of three,
Index view is to store the results of a query database objects,
Transaction internal fault can be divided into expected and unexpected, unexpected affairs internal fault include: arithmetic overflow fault, concurrent transactions deadlock failures, violated some integrity constraints caused by the fault,
In t-sql, can achieve points according to different types of data function is the CASE,
Used in the database transaction profile design phase, the read and write the original language to express the access operation,
Use TOP limiting the result set, if you need to get the query result first n % rows of data you need to use the TOP n percent option,
Common table expressions can be the result of a query set to specify a temporary naming names, the name of the result set is called a common table expression, named after the common table expressions can be reference for many times, syntax format is: WITH common table expression identifier (list 1... N) AS (SELECT statement),
Call a stored procedure, the DEFAULT used to insert a DEFAULT value in a column,
Delete the stored procedure: DROP the PROC,
When using REVOKE commands recycling user permissions, if want to will the user switched to other user permissions. Also, the command need to add in the CASCADE option,
Establish the CLUSTERED INDEX view called INDEX view, defined statement: CREATE UNIQUE CLUSTERED INDEX INDEX name ON view name,
Data from the operational environment is transferred to the ETL tool of data warehouse process used usually need to complete processing operations including extraction, conversion and reproduced,//ETL is the main technology to realize data integration, which fill the update data extraction, data warehouse, data acquisition process of reproduced,
Data warehouse is a feature: subject-oriented, integrated, not easy and change over time,
Export data of data warehouse and materialized view maintenance policies, found only when the user query data has expired to update to update strategy called delay maintenance policy,
Generally divided into data warehouse, metadata technology and business metadata, metadata
Database performance optimization, storage optimization including establishing materialized views and aggregation,
When a database system failure occurs, fault tolerance is after the restart the database, to cancel all uncommitted transaction, redo all committed transactions,
In cloud computing, SaaS refers to software as a service,
Usually includes software as a service, platform as a service, infrastructure as a service,
Clustering algorithm by processing the data objects are generally no tags, so the general also called unsupervised clustering method of study,
Three-tier browser/server structure of the database application system is by the browser, the database server and application server of three parts,
To save the query results to the new table: SELECT queries list sequence INTO & lt; The new name of the table & gt; FROM the data source
Local data model transparency between distribution model and partial concept,
Database implementation and deployment phase mainly include: to establish the database structure, data loading, transaction and the application of coding and testing, system integration, testing and commissioning, system deployment,
If you want to allow a user to give his privileges in re-delegate its other users, so need to add in the GRANT statement WITH GRANT OPTION OPTION,
IDEF0 requirements modeling method has two kinds of arrows and rectangular elements,
In IDEF1X data modeling method, orthogonal rectangular boxes represent separate entity set, rounded rectangle box used to represent a subordinate entity set,
IDEF1X, many-to-many connection is called a non-deterministic,
Modify the DB1 DATABASE structure, the use of the statement is: the ALTER DATABASE DB1...
The source of the login account has two kinds: one kind is SQL SEVER themselves responsible for authentication login account; Another kind is log in to the SQL SERVER Windows network account, can be a master account or user account,
Increased redundancy column tuning method is refers to the increase in multiple tables with the same semantic columns, when he used to query to avoid join operations,
Increase the redundant column is derived, compute a column (for example: the total price)
In classification forecasting task, need to use the data generally include training set, test set and validation set,
Firewall technology is used to maintain the security of database application system of the network environment,
Will often use multiple rows in the table according to its public column values stored in together, this table storage method for gathering,
DIFFERENTIAL backup SQL keyword: DIFFERENTIAL
Dense index: if the data in the file each lookup code value corresponding to an index in the index file records,
Provides four ranking function SQL SEVER:
RANK function returns a result set data in each row in each partition in the rankings, and each partition in the rankings, and within each partition number starting from 1, but the middle numerical rankings may have continuously;
DENSE_RANK result set partition the bank of China, there is no any gap in rank;
ROW_NUMBER: the serial number of the result set partition adept, each partition of the first line from the start, the ORDER BY clause can be set in a specific partition only ROW_NUMBER ORDER for line distribution,
NTILE: the orderly division of line distribution to specify the number of groups,
The SQL SEVER, delete the DB1 database User U1: DROP the User U1,
The SQL SEVER, delete the DB1 LOGIN account U1: DROP the LOGIN U1,