"In-memory database has entered a new stage. Cloud native, persistence, and converged computing are the three major characteristics." On December 13, Li Feifei, head of Alibaba Cloud Database, said at the Cloud In-Memory Database Summit, after years of verification of applications in Alibaba's economy , Alibaba Cloud's in-memory database Tair, launched at the beginning of this year, has entered the stage of large-scale sharing of technology and hardware dividends to users on the cloud. Tair is a self-developed cloud-native memory database developed by Alibaba Cloud. It has high throughput and low latency, and provides the endurance capability of RPO=0 to make data online in real time.
Tair is widely used in the core system of Alibaba's economy and has stably supported Tmall Double Eleven for 12 consecutive years, with peak visits of 1.5 billion per second. In complex application scenarios, Tair continues to make technological innovations. The innovative Tair hotspot hashing solution for hotspot scenarios makes Internet applications similar to e-commerce, social networking and other frequently occurring hotspots, which greatly reduces application development costs. , And save at least 2/3 of the resources reserved for hot scenes for applications; in the scenes of massive hot data storage such as product information, Tair introduces higher density and lower cost persistent memory, saving the overall TCO of the cluster up to 30% At the same time, throughput and latency can meet the needs of scenarios, and the solution can be reused in online data access scenarios in industries such as e-commerce and online education.
After going online at the beginning of this year, Tair has attracted many users with higher performance and enterprise-level functions. It is reported that Pumpkin Movie uses the enterprise-level functions of the Alibaba Cloud in-memory database Tair to achieve data synchronization in multiple regions, bringing a more extreme experience to 30 million end users.
In the cloud era, in-memory databases will continue to develop in low-cost, flexible, and enterprise-level functions. With the development of database memory technology and the popularity of persistent memory hardware, in-memory databases will continue to reduce costs to reflect their cost-effective advantages; the rapid development of cloud computing makes applications have high requirements for the elasticity of basic services. High-throughput and low-latency scenarios have higher requirements for flexibility; in-memory databases will undertake more calculations, and enterprise-level functions such as computing push-down memory, global distributed databases, and recovery at any point in time will become the basic requirements of users for in-memory databases .
According to a market research report released by VerifiedMarketResearch in 2020, the global in-memory database market will grow at a compound annual growth rate of 19.65%.
According to SumoLogic's 2020 survey report on 2100 companies, the in-memory database Redis has become the most popular database on Amazon AWS. Yiftach Shoolman, the chief technology officer and co-founder of RedisLabs, said on The Register last year that the company hopes to continue to transform Redis's view from caching to a "real database."
Tair, the Alibaba Cloud memory database, is compatible with the Redis interface on the one hand.
In the past 12 years, the transition from caching to in-memory database roles has been gradually completed. In addition to using Tair's cache mode, more and more core applications also use Tair as the final storage solution for data, and sink and integrate calculations into Tair to simplify The overall application architecture. In the future, Tair will continue to make efforts in cost, flexibility, computing, and enterprise level to provide more popular services to help 1,000 enterprises simplify application architecture and make data online in real time.