In 2023, the data center needs to be "distributed"
In April 2022, the first edge computing computing center in Chengdu - Yunyou Technology Kunpeng Ecological Edge Computing (Chengdu) Computing Center was put into operation;
In August 2022, the distributed small data center module of Zhangjiajie Siji Branch located in 110kV Wulingyuan substation will be put into commercial operation;
On November 10, 2022, the first pilot project of Hangzhou's innovative demonstration "985 Project" distributed data center business model was officially unveiled in the distribution room of Hangzhou Qianlong Electrical Appliance Co., Ltd
In the past year alone, the application projects of data centers at the edge of China have sprung up like mushrooms.
On the other hand, in the context of "data from the east to the west", more and more large and super large data centers have sprung up in 8 nodes and 10 clusters. An invisible computing net has begun to weave. The data center is gradually developing towards two polarization - edge side, small and micro data center; In clusters, large and super large data centers.
"Conflict" between "physical" and "logical"
On the one hand, the data center construction shows a trend of "polarization"; On the other hand, users want a "logical" completely unified virtual resource pool to achieve faster and more convenient deployment of business. The "conflict" between "physical" and "logical" is in front of us. How can we better realize the link between data centers, achieve data and computing, and flexibly and quickly schedule and communicate among multiple data centers? The distributed data center may become the "optimal solution" to the above problems.
The distributed data center is actually called the distributed cloud data center. Its essence is based on cloud computing technology. Thousands of servers are formed into a huge virtual host through virtualization technology to achieve unified scheduling and management of data and computing power, avoid repeated investment, and make better use of hardware, software, and network resources.
"Physically decentralized construction, logically unified management" ——This is the simplest statement to describe the distributed data center. At the "physical" level, the data center can be built anywhere, even underwater and in space; At the "logic" level, virtualization technology is used to make the data centers built around work better together, so that users can simply use the data center to conduct business like a home computer.
It is easy to see from the development of the data center industry at this stage that the demand of users for computing power is growing. Of course, the most ideal state is to build a data center that is large enough to meet the demand of users for huge computing power, but this situation is only "ideal". In reality, users need to build multiple data centers in the same place, or even in different places, to meet their computing power, as well as the calculation and storage requirements of different data (temperature and cold data, thermal data).
At this time The distributed architecture becomes the key to solve the "conflict" between "physical" and "logical".
The distributed data center can be divided into two categories from the application level—— Single data center distributed architecture, multi data center distributed architecture.
Single data center distributed architecture
This kind of architecture is mainly integrated by distributed computing, storage, network and other distributed technologies. It is mainly used in the single data center. Distributed, virtualization and other technologies are used to maximize the use of servers and improve efficiency.
Multi data center distributed architecture
This kind of architecture is mainly applied to multiple data centers in different geographical locations to achieve unified management of multiple data centers. Its core technology is still cloud computing technology and WAN interconnection technology. Build multiple data centers into a distributed "virtual resource pool" integrating computing, storage, network, etc. This resource pool can integrate and manage data centers in different regions in a unified way, and even provide disaster recovery for each other. It has the characteristics of improving enterprise efficiency, reducing the cost of renting data centers, and reducing security risks.
Whether it is a single data center distributed architecture or a multi data center distributed architecture, in the author's opinion, when the data center enters a critical period of digital transformation, the distributed architecture becomes the core of the edge data center, and whether the "East Data West Computing" project can "live up to reality".
Why distributed architecture?
Why is the distributed architecture an indispensable support in the development process of the "polarization" of the data center? Because based on cloud computing related technologies, data centers with distributed architecture can solve the "conflict" between the "physical" and "logical" levels of traditional data centers, so that multiple data centers located in different locations can form a large "computer". In addition to this advantage, the distributed architecture has three advantages over the traditional data center architecture.
1、 More stable
With the advancement of enterprise digitalization process, its business has become increasingly dependent on the data center. Once the data center is "down", the loss to the enterprise is incalculable. If the distributed architecture is adopted, enterprises can easily synchronously backup data in multiple, remote data centers. On the one hand, the risk of power interruption or downtime is avoided to ensure business continuity;
On the other hand, due to different geographical locations, it can also avoid the impact of "natural and man-made disasters" such as earthquakes, floods, and operational errors on the business of enterprises. For example, on October 15, 2022, due to the fire in SK C&C Banqiao Data Center, Kakao's chat software Kakao talk and other applications were interrupted, and a large number of South Korean users were affected; On December 18, 2022, Alibaba Cloud's Hong Kong machine room node failed, causing the websites of key infrastructure operators such as Macao Financial Authority, Galaxy Macau, Lotus Satellite TV, Macao Cement Factory, and other takeout platforms such as Omi and MFood, as well as APP such as Macao Daily to be unable to access and use normally for a time...... If a multi data center distributed architecture is adopted, It can minimize the impact on users after disasters.
At present, the financial industry is the most mature application of distributed data center. As we all know, under the supervision and its own requirements for data security and business continuity, the multi center distributed architecture scenario evolves from the "local disaster recovery center" to the "dual active" architecture (i.e., production data center, local disaster recovery center, and remote disaster recovery center). With the development of business, it may evolve into the "multi center distributed multi active" architecture in the future.
At a time when most banks have regarded data as assets, data asset management and data potential stimulation through distributed technology have become an important means to enhance the overall competitiveness of banks.
2、 Cheaper
The distributed data center allows users to select not only one service provider, but also more focused on the development of their own business, without too much attention to the details of bottom layer technology operation and maintenance, machine room management, etc. On the one hand, it helps users save the cost of operation and maintenance personnel, on the other hand, it enables users to deploy and purchase on demand. For key core data, It can be purchased and managed separately from non key data, and can also allow users to spend every penny on "cutting edge" to achieve real cost reduction and efficiency increase.
In theory, users are not limited to the services of a single data center operator. For example, you can use a virtual server in a data center in Switzerland, a server in a data center in Britain, and a server in a data center in Russia.
Moreover, in the context of "counting from the east to the west", at this stage, the "calculation cost" of the western region is obviously lower than that of the eastern region. Some non real-time data processing, such as video rendering, machine learning, large-scale data storage, etc., enterprises can choose to process in the western region. Through distributed technology, these data and computing power can be more smoothly scheduled, and the cost of enterprise leasing services can be further reduced.
3、 Faster
As we all know, in the Internet era, the speed is extremely high. Sometimes business data flows need to be calculated in milliseconds or even smaller time units.
For example, in stock trading. Even millisecond delay will bring considerable competitive advantage to competitors, who can buy and sell stocks earlier by taking advantage of this advantage. For this reason, securities dealers choose to use data centers in Frankfurt and London near the main exchanges.
For another example, with the gradual popularization and application of automatic driving technology, the data center needs to be quickly calculated at the edge to ensure that the automatic driving vehicles can handle emergencies and complex road conditions in real time, and ensure the safety of passengers and surrounding people and things.
At this time By adopting distributed technology, the data center can be closer to the user side and the edge side, so that data and computing power can be better scheduled, Compared with the traditional centralized management, the circulation and calculation between nodes are faster and more efficient.
Four major challenges cannot be ignored
Although the distributed data center has the advantages of being more secure and helping enterprises reduce costs and increase efficiency, as far as the existing technology is concerned, there are still four major challenges that the distributed data center needs to face if it wants to mature applications.
1、 Computational scheduling
The management of computing resources is a difficult problem at present. For data centers located in different locations, how to efficiently schedule computing resources, realize data in different geographical locations, and unify management when dealing with business data flow emergencies will become a major challenge for distributed data centers in computing scheduling. Using the distributed data sharing governance model to "integrate" centralized and distributed data governance and achieve coexistence is one of the more common and mature means at present.
2、 Cloud network integration
The core foundation for realizing the distributed architecture of multiple data centers is the network, and doing a good job in cloud network integration is also the key to the success of "calculating data from the east to the west". As mentioned above, the realization of "logical unified management" determines whether the distributed data center can be applied.
Based on China, the main force to promote cloud network integration should be the three major operators. At present, and for a long time in the future, the three major operators will still be the main force of China's network construction.
3、 Distributed Storage
How to realize data collaboration is one of the main challenges currently facing. With the high coverage of business, the importance of collaboration among local data centers is increasing. However, due to the difficulty of distance and scale, the network broadband between local data centers cannot ensure real-time data synchronization, which has a certain impact on the consistency and integrity of data and the continuity of business.
In the middle of 2022, Gartner announced the maturity curve of storage and data protection technology in 2022. The software defined storage infrastructure became the only revolutionary technology of the year, and the priority rating of distributed file system, object storage, and container native storage was "high".
obvious, Distributed storage will become the next market wave for storage manufacturers 。
4、 Data security
Safety has always been one of the most concerned topics of enterprises. Although distributed data centers are more flexible than traditional data centers, how to ensure the security of data assets from both physical and virtual levels has become a major challenge.
Under the distributed architecture, data will flow among multiple data centers. After the architecture of the distributed data center becomes more complex, the security challenges are also increasing. In the process of circulation, how to build firewalls, build security protocols, and ensure data security has become a major challenge in the application process of distributed architecture. It is reported that a complete and unified security product solution has not yet been formed.
With the development of the data center industry in recent years, it is an indisputable fact that the data center is gradually moving towards "polarization". Distributed technology will play a vital role in the construction and application of small data centers on the edge, as well as in the overall development of large and super large data centers in clusters.
At this stage, its technical advantages and challenges are obvious. With the development of 5G, Internet of Things, industrial Internet and other technologies, the author believes that 2023 will be a breakthrough year for distributed data center technology.
