Under the new productivity revolution, Amazon Cloud Technology has more than 600 different computing instances to meet the computing power requirements

4182a1acb6d24d968861cf587608494a.pngRecently, the annual Amazon Cloud Technology China Summit was held in Shanghai as scheduled. At this summit, you can really feel the Day one concept of Amazon Cloud Technology, and also made a detailed interpretation of AI innovation. 

"AI innovation" makes the demand for computing power blowout and the complexity of operation and maintenance increases

With the development of AI large-scale models, the scale of large-scale model applications has grown explosively, which has accelerated the advent of the future computing era and triggered a new productivity revolution. Due to the strength of data processing capabilities, users not only need more cost-effective hardware solutions, they also need cloud vendors to provide highly elastic scaling. At the same time, the huge increase in computing power has greatly increased the cost of enterprise cloud operation and maintenance. How to reduce the complexity of cloud operation and maintenance is also a difficult problem facing enterprises.

In order to meet the demand for elastic computing power, Amazon Cloud Technology provides more than 600 different computing instances. The processor, storage, network and various peripheral services can be well combined with computing to form a rich and flexible computing environment in the form of building blocks. Computing instance resources to meet various computing power requirements.

It should be known that from the storage level, when the data scale reaches the PB or EB level, the data storage method is very important. Users' data storage can be divided into four types: hot, warm, cold, and frozen. Each type has different storage costs and performance, and needs to be balanced. Therefore, Amazon cloud technology provides up to eight storage levels, including high-performance computing, machine learning and general computing, and enterprises can choose the most reasonable storage level according to their needs.

The most representative case is Weta Digital, the production company of "Avatar". When developing "Avatar 1", they built a computer room covering an area of ​​10,000 square feet and built 4,000 physical servers and 35,000 processors. core. However, when developing "Avatar 2 - The Way of Water", their original cluster could no longer meet the needs. Therefore, Amazon Cloud Technology provided Weta Digital with a series of services ranging from infrastructure computing resources to cloud production stacks and machine learning stacks. Weta Digital used a variety of computing instances including GPU instances and Spot instances to provide powerful Scalability and excellent cost performance.

However, for Amazon Cloud Technology, it is not enough to provide enterprises with a variety of chips and highly scalable resources. Amazon Cloud Technology also provides Serverless for enterprises to reduce the complexity of cloud operation and maintenance. Since Amazon Cloud Technology released Lambda in 2014, it has successively released more than 105 new serverless technologies, including Lambda SnapStart, which solves the cold start problem of Java applications, and Application Composer, a visual editor. Users can choose different elasticity and calculation granularity according to the type of application.

Fully managed Serverless allows developers to focus on business development rather than infrastructure operation and maintenance, and also allows the system to easily achieve rapid scaling. To this end, Chen Xiaojian gave a special case of Second Dinner creating the best mobile game of the year "Marvel Snap" at the summit. "Marvel Snap" was called by Second Dinner's vice president of engineering "this is the smoothest and most successful game we have ever operated, precisely because we chose Amazon cloud technology".

Second Dinner uses Amazon Cloud Technology's Serverless architecture to develop, build and run the Marvel Snap game. The entire game does not use any EC2 computing instances or containers, but is implemented by an event-driven architecture. This approach not only helps users save costs, speed up application development, but also reduces security risks. The serverless background architecture of Amazon cloud technology provides it with complete functions, security and resource management. Ultimately, Marvel Snap launched without any incidents of backend errors, which is unprecedented in the gaming industry. In addition, the developers of Marvel Snap have already achieved a stress test of 140,000 requests per minute through Serverless in the very early testing stage of the system. You must know that this is unachievable in the era of cloud hosts, and this is possible in Amazon Cloud Technology The "impossible" was realized with the help of

Guess you like

Origin blog.csdn.net/2201_75638547/article/details/131472621