Continuous construction of Data and resources with efficiency-enhancing computing

Fancy Wang 2310 2020

Some of the following content comes from Broadband China

1. Data Center cluster to scale, intensive development

Globally, as new technologies such as 5G, cloud computing and artificial intelligence accelerate to the ground and data traffic enters a new phase of explosive growth, data centers will accelerate toward super-scale, green energy-saving direction. According to Ericsson’s forecasts, the 2024’s global mobile data traffic will reach 131.5 exabytes per month, 25.4 times higher than in 2015(5.2 exabytes) .

In order to make full use of data, data center scale efficiency, greatly reduce the cost of business deployment and maintenance, large, super-large data center is favored by operators, Internet and other data center providers. Since 2013, the number of super large data centers worldwide has tripled, with Amazon, apple, Google, Facebook and Microsoft leading the way, and the largest server on a single campus now exceeds 300,000, many large campus servers range in size from 20,000 to 100,000. Gartner predicts that by the end of 2020 there will be more than 600 large data centers worldwide, accounting for 47 percent of all data centers and servers.

At the same time, data centers are big consumers of energy, and global data centers account for 3 percent of the world’s electricity consumption, and by 2025, global data centers will use, more than 10 billion dollars of electricity at current prices, compound annual growth rate of 6% . In order to reduce the energy consumption and improve the operating efficiency of data centers, governments and related enterprises around the world have taken a series of measures to promote data centers more energy-efficient.

From China’s point of view, through the introduction of policies such as the guiding opinions on strengthening the construction of green data centers and the rolling update of the national guidelines on the application and development of data centers, we will actively guide the construction and distribution of data centers in a rational way, continue to move toward large-scale and intensive, effectively improve the national data center, the energy efficiency level. At present, the power efficiency values (Pue) of super-large and large data centers are 1.63 and 1.54, respectively.

It is expected that by 2020, the energy consumption can basically be comparable to the international advanced level, and the Pue of newly-built large and super-large data centers can reach less than 1.4. On the other hand, companies are adding, and fast technological innovation is driving data centers to reduce energy consumption. Data Center providers realize the green development of data center by applying advanced technology products such as liquid cooling service, server, high density integrated IT equipment, high efficiency refrigeration system, city power supply, DC power supply, etc. . For example: ALIBABA’s Panama, power supply is to address the Data Center DC power demand of the latest generation of power products, can effectively improve the data center power efficiency of 3% , reduce the total investment cost of power supply and distribution by 20% .

2. The computing mode is reformed from the traditional isomorphic computing to the heterogeneous computing.

Global mainstream vendor computing architecture to heterogeneous development. Diverse application scenarios drive heterogeneity and accelerate the arrival of the age of structured computing. Driven by new technologies, such as 5G, artificial intelligence, big data, and the Internet of things, Internet applications are increasingly diverse, and the demand for massive computing is growing exponentially, the demand for computing power outpaces the growth of Moore’s law and is characterized by differentiation and individualization, the performance development of homogeneous computing (also called general-purpose computing) , which is represented by traditional CPU chip, can not keep up with the demand of application.

The global mainstream factory, the business through the software, the cloud and so on way, deployment heterogeneous computing open source ecosystem, reduces the heterogeneous, the computation popularization use complexity and the use cost, speeds up its popularization application. For example, Intel introduced a unified software platform ONEAPI, through CPU, GPU, AI, FPGA and other computing units, according to user needs for flexible combination, to provide users with cross-architecture, cross-platform group, combined solution.

Accelerating the formation of heterogeneous computing ecosystem in China. Chinese chip manufacturers integrate special-purpose computing units into general-purpose chips, and there are both traditional general-purpose computing units such as CPU and high-performance special-purpose computing units in their computing architectures. Heterogeneous computing through CPU + GPU, CPU + FPGA, CPU + Asic and so on, is playing more and more important role in carrying application, it is becoming the mainstream of the industry gradually, and this trend may continue to accelerate.

For example, Huawei builds a heterogeneous computing hardware resource cloud management platform to assign the best heterogeneous hardware resources for different computing tasks. As the open source ecosystem of heterogeneous computing takes shape, new businesses, new applications continue to pursue computing power and optimal energy efficiency, and the open, pervasive heterogeneous computing ecosystem will accelerate its formation.

3. Cloud edge depth fusion is an important direction for the future.

The deep convergence of cloud edge is based on the parallel driving of business requirement and technology innovation, which brings about profound changes in network architecture. With the coming of the 5G era, edge technology is regarded as one of the key innovation technologies to change the mode of information communication and information service, which is more suitable for real-time, short-period data processing and analysis. Countries around the world have reached a consensus on the cloud-side synergy route.

The world’s leading manufacturers have launched a series of cloud services and solutions in collaboration with the industry ecology. Microsoft, Google, Amazon and other companies have actively deployed cloud-edge collaborative development, amazon’s AWS Greengrass, Microsoft’s Azure IoT Edge, Google’s TPU and Cloud IoT Edge. IDC’s data age 2025 report predicts that by 2025,50 percent of the data will be analyzed, processed and stored on the edge of the network, while traditional cloud computing, which relies on centralized computing resources, has already, can Not meet the exponential growth in the number of devices and the deployment needs of “large connectivity, low latency, large bandwidth” applications, only cloud computing and edge computing in close coordination, can better meet the needs of a variety of scenarios, real-time and fast data storage, intelligent processing and prediction analysis.

China’s cloud network integration to accelerate the penetration of the industry. Cloud network convergence is based on business needs and technology, innovation brought about profound changes in the network architecture, making the cloud and the network highly collaborative, convergence and complementary. Currently, cloud-to-network fusion is mainly used in cloud-only lines and SD-WAN scenarios.

With the development of cloud-to-network Convergence, the field of cloud-to-network convergence will gradually extend from simple interconnection to “cloud + network + business” , and integrate with specific enterprise applications, providing obvious industry attributes and satisfying users’ poor performance, the integration of downlink and ICT services makes the integration of cloud networks, integration products and basic service capabilities more closely, and finally realizes the reasonable allocation of enterprise computing resources and dynamic adjustment of service resources, provide customized cloud computing services that users need.

For example, China Telecom has independently developed an edge computing platform based on a distributed open platform, which supports fixed/mobile network access, third party capability/application flexible deployment, and unified and open edge capability, it can be used in industrial internet, high definition video, vehicle network and other fields to solve the problems of low delay and data security in industrial control.

About Us

About Shenzhen Speed Technology Co.,ltd, we are a research and development switch manufacturer located in Shenzhen, Guangdong. We started in 1997 with a variety of product lines, supporting OEM, ODM customization.