The present situation and challenge of the infrastructure of Yiwen parsing database

Status of database infrastructure

At present, information system has been an important part of supporting modern enterprise operation , There are dozens or even hundreds of different in the enterprise IT Application system , Support the service channel of the enterprise 、 Customer marketing 、 Product business processing 、 Management and decision-making , Most application systems revolve around database development and operation . There are generally more than ten core application databases and dozens of general application databases , Multiple databases and different versions at the same time 、 Databases of different specifications are running .

for example , A regional bank owns 6 set Oracle RAC, Per set 3TB The amount of data supports the core accounting , Other general production operations use more than 100 set 、 Capacity from 2TB To 100GB No wait MySQL; In addition to the core HIS System USES 1.4TB Data quantity Oracle RAC As a support , Other EMR 、 Electronic bills 、 Medication guidance, etc 5 An application system , use Oracle Database as support , Capacity from 80GB To 1TB Unequal ;PACS、LIS、 Waiting in line in the outpatient department 17 Adopted by systems SQL Server Database as support , Capacity from 80GB To 800GB Unequal ; Paperless medical records are used MongoDB As a database support , Capacity of 400GB.

at present , The infrastructure of these databases mainly adopts the traditional integration architecture 、 Virtualization deployment mode and database all-in-one machine , The positioning of each option is outlined below ：

The above banks 、 Medical customers also adopt traditional architectures based on physical servers and medium - and high-end storage , And the virtualization based architecture supports the database operation of different business requirements . The following will give a more detailed analysis of the advantages and problems of the three mainstream architectures .

Advantages and disadvantages of mainstream database infrastructure solutions

1、 Based on traditional integration architecture （ The physical machine + Centralized storage ） Database solutions for   

In the actual deployment environment , To ensure high availability of the database 、 High performance 、 Features of large capacity , At the same time, it can avoid that the service cannot be accessed due to a single point of failure of the server , Generally, the database application instance and database separation deployment mode will be adopted , use 2 One or more servers serve as database instance servers , adopt SAN The network is connected to the medium and high-end high-capacity SAN The storage device , Data access services . The details are as follows: ：

The advantages of this deployment are ：

1. Ensure that the database instance is highly available （ Through active and standby 、 Hot standby or cluster technology ）.

2. use SAN High end storage technology , Ensure high performance storage 、 The large capacity .

3. Utilizing storage devices RAID characteristic , Ensure data security 、 complete 、 reliable .

4. Through the server 、 Hierarchical construction of storage devices , Realize the high scalability of the system , Meet the needs of application deployment .

The configuration and cost estimate of the scheme are as follows ：

In addition to the purchase cost of the product , Other services and maintenance architectures are referred to as follows ：

The system integration deployment cost is based on the purchase cost 12% Calculation ,1388000 * 12% = 166000

3 The annual maintenance cost is calculated as per 8% Calculation ,1388000 * 8% * 2 = 222000

total cost ： about 1770000

During the construction of the data center with storage as the core , This method is widely used , It strongly supports the user application system 、 Database deployment and business development . Now many users' legacy systems , There are a large number of deployment cases .

There are two main problems with this model ：

First, according to Moore's law ,IT System, especially server CPU、 Memory, etc. , Performance doubles every two years . The performance of network and disk , But it can not be improved at the same speed . In this way, network and disk become the bottleneck of system performance . Mainly reflected in ：

1. Storage level , The amount of data is increasing , bring IO bottleneck ; Uneven data distribution caused by long-term data operation , There is IO hotspot .

2. At the network level , Insufficient transmission bandwidth , Unable to quickly transfer a large amount of data to the server for processing .

3. Server level , Received too many invalid data to be processed , The advantages of high-performance processors and memory cannot be realized .

In response to this problem , Some database service providers conduct in-depth R & D and integration based on software and hardware , Launched the all-in-one machine of extreme performance database .

The second is the deployment mode based on physical machines , In case of equipment failure 、 System upgrade 、 Capacity expansion 、 Maintenance, etc , This can lead to planned and unplanned downtime , This will affect the normal operation of the user's business .

therefore , With the development of virtualization technology , Many users in VMware And so on VM Deploy database , Through virtualization technology HA、vMotion、DRS、 snapshot 、 Cloning and other technologies , Ensure virtual machine VM And database system have high availability , Reduce database and IT Downtime of the application system .

2、 Database virtualization deployment scheme

The following figure is a typical topology diagram of deploying a database on a virtualization platform , Centralized storage including the core 、 Storage switches and multiple x86 The server . By adopting VMware Virtualization technology , stay x86 Create a virtual machine on the server VM, In the virtual machine VM The internal installation operating system runs Oracle、MySQL、SQL Server Wait for database management system , The data of the database is stored in centralized storage .

The advantages and characteristics of this architecture are ：

1. Take advantage of virtualization technology , Improve availability , Can be dynamically migrated , It can support high-end （HA）, The system can be upgraded without interrupting user applications .

2. By taking advantage of snapshots 、 Cloning and other technologies , When there is a problem with the upgrade , Rollback is supported , Avoid system failure shutdown , adopt DRS Dynamic provisioning , May enhance IT Adaptability of the system to applications and businesses .

3. Support online expansion x86 The server , You can add computing resources as needed .

The configuration and cost estimate of the scheme are as follows ：

The cost of system integration deployment is based on the purchase cost 12% Calculation ,1198000 * 12% = 143000

3 The annual maintenance cost of the hardware is calculated as per 8%, Software on an annual basis 22% Calculation ,1088000 * 8% * 2 + 110000 * 22% * 2 = 222400

total cost ： about 1560000

The main disadvantages and problems of this architecture are ：

1. This architecture focuses on centralized storage , Limited by the number and performance of storage system controllers , The overall performance of storage is limited , Capacity can only be expanded vertically , Unable to improve storage performance through capacity expansion , With the increase of computing resources , virtual machine VM The average performance of is declining , This will affect the virtual machine VM And database performance .

2. Inconvenient operation and maintenance management , The whole system is divided into several IT level , Maintenance and changes need to be carried out in multiple equipment maintenance systems （ Include Web、 graphics 、 Command line ） The deployment shall be coordinated by relevant experts and technicians , High labor costs .

3. use VMware And other virtualization technologies enhance the flexibility of the system , But it also adds IO route , Reduced system performance , Unable to meet some high performance requirements VM And database operation .

3、 Database all in one solution

In database architecture design , There are mainly Shared Disk,Shared Nothing Two structures .

Shared Disks The architecture and Shared Nothing Each architecture adapts to different scenarios , With Oracle The database service providers represented by the company combine the characteristics of these two architectures , Provides a hybrid database architecture , Can effectively solve the conflict between the two , Draw on the advantages of both architectures ： Both can satisfy OLTP High concurrency 、 High availability features ; Can satisfy again OLAP Large data processing requirements . Its structure is as follows ：

Based on this hybrid architecture ,Oracle The company has launched an all-in-one machine for extreme performance database Exadata, The main features are ：

• It has been completely pre installed by the manufacturer's software and hardware experts in the factory , Open the box .

• Use the most advanced software and hardware technology in the industry , And bring these technologies to the extreme .

• Solve the performance bottleneck in traditional system configuration .

• Provide comprehensive tools and means , Simplify management and operation and maintenance .

This architecture can improve database performance , Simplify the deployment of database systems , But the disadvantages are also very obvious . High procurement and operation and maintenance costs , inflexible , Only proprietary databases are supported , And they are all special systems , The architecture is closed and not open , Unable to meet the needs of cloud transformation .

The configuration and cost estimate of the scheme are as follows ：

（ notes ： It doesn't contain Oracle DB Permits and RAC License fees ）.

in addition , Many domestic companies have launched similar all-in-one database machines , Such as Wacker company QData All in one database machine , Yunhe enmo company zData All in one database machine , Tianji data company PBData All in one database machine . They also have out of the box , Simplify management and operation and maintenance , But in the aspect of software and hardware optimization , and Oracle There is still a gap between the original all-in-one database machine .

Summary of database infrastructure challenges

At present, with the acceleration of enterprise digital transformation process , Enterprises' demand for databases is increasing , Types are increasing , Agility requires more , At the same time, simpler architecture and more reasonable cost of ownership are required to cope with the rapid growth of business , But summary The main problems of the above mainstream programs include ：

• The cost of all-in-one database machine is high , Complex operation and maintenance , Poor elasticity , Cannot support multiple databases .

• Physical server plus centralized storage , At cost 、 Poor maintainability and agility .

• VMware Add the scheme of centralized storage , In terms of performance and delay, it still lags behind the bare metal and database all-in-one solutions , At the same time, the storage scalability and overall cost still can not better meet the needs of users .

meanwhile , Financial and other industries have also emerged New policies and technical routes need transformation ：

• At present, the financial and other industries are facing the needs of localization and transformation , Facing a new problem of hardware and software adaptation , Greater challenges to database infrastructure .

• At present, the container and K8s Cloud native architecture based on cloud puts forward higher requirements for the agility of infrastructure .

in summary , The user database scenario requires Meet the performance 、 stability 、 agility 、 Simplicity and even a series of new solutions for the original requirements of Xinchuang and cloud .SmartX Based on hyperfusion 、 Distributed storage 、 Yunyuan survival storage and other core products , For different user scenarios , It provides a series of database cloud infrastructure solutions , The database resource pool is not only easy to deploy and operate 、 The public cloud features flexibility and agility , At the same time, the performance experience consistent with the traditional architecture .