Shulou(Shulou.com)12/24 Report--

Thanks to CTOnews.com netizens Wu Yanzu in South China for the delivery of clues! CTOnews.com, Nov. 29, in the Amazon "AWS re:Invent 2023" event launched in Las Vegas today, Peter DeSantis, senior vice president of Amazon's computing department, introduced its three cloud services. CTOnews.com collated the relevant information as follows.

Peter DeSantis first introduces Amazon's new version of Aurora service, which is a serverless cloud database service compatible with MySQL and PostgreSQL. The biggest improvement is the ability to expand capacity online, which mainly depends on virtualization technology. Amazon has developed a virtual layer called Caspian, in which Caspian Heat Management System manages how physical hosts allocate virtual RAM to databases.

CTOnews.com Note: the serverless version is "serverless computing", which uses the cloud to provide computing services directly. Under this architecture, developers only need to write code and upload, and the cloud platform will automatically prepare the corresponding computing resources, complete the calculation and output the results, thus greatly simplifying the development, operation and maintenance process.

It is reported that each physical host used by Aurora can configure the RAM of 256GB. Although the Aurora database opening service can use 256GBRAM, the RAM of the physical host will not be fully occupied. The Caspian management software will arrange the appropriate physical RAM capacity according to the configuration required for the operation of each Aurora database service. Therefore, each physical host can carry multiple Aurora database services, thus greatly improving the computational efficiency. Full use of computing resources.

However, under the above mechanism, if a physical host requires more than 256GB of RAM, the user must manually transfer to another host due to the physical limitations of RAM. DeSantis pointed out that the current "serverless computing technology can not make this function fully automatic", so Amazon uses database partitioning (Sharding) technology to allocate a single database to different hosts to achieve a "side level" to expand RAM capabilities.

DeSantis said Amazon had developed a lightweight data request routing layer (Request Routing Layer) with rapid response expansion, which claims to be able to "slice and reorganize the database across regions and ensure that the database is still reliable".

Combining the above two technologies, DeSantis pointed out that Amazon AWS will finally be able to launch a truly serverless Aurora database service this year, namely "Amazon Aurora Limitless Database", which claims to be "not limited by the capacity of a single database, and can automatically expand the host configuration according to user needs without affecting the operation of the database."

▲ source Amazon in addition to Aurora Limitless Database, Amazon today also launched a serverless version of ElastiCache cloud caching technology.

DeSantis said that the synchronization of data in distributed computing systems depends on synchronous timing, and the industry usually uses Wall Clock (real world time), Logical Clock (logical time) and TimeKeeper servers, but Wall Clock will make timing synchronization unreliable due to clock differences.

To provide accurate timing, Amazon claims to have launched its Amazon Time Sync service in 2017, which combines satellite and atomic clock information references to provide timing with an error of only 1 millisecond (millisecond) with the Universal time Coordination (UTC).

However, DeSantis pointed out that a time error of 1 millisecond represents a total delay of 2 milliseconds, or "500transactions per second", which is "too slow" for cloud cache synchronization.

Therefore, Amazon has designed a standard 48U cabinet size distributed timing network based on the Nitro network acceleration chip, equipped with independent power support and a physical atomic clock, which claims to reduce the synchronization error from milliseconds to microseconds (microsecond), reducing the delay of the ElastiCache Serverless p50 to 500 microseconds.

Related reading: "Amazon launches new AI quantum chip Trainium2 and Graviton4 processor to cope with Microsoft competition."

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.