From e-commerce websites to sophisticated ERP platforms, database systems operate silently behind the scenes to capture and direct the complex and continuous flow of data in today’s digital world. And as the world moves towards the cloud, 75 percent of all databases are expected to be deployed or migrated to a cloud platform by 2022, according to analyst firm Gartner. This reality is even more significant in Asia-Pacific where the cloud computing market is expected to skyrocket,exceeding US$250bn in the region by 2023..

Confronted with growing demand by enterprises for cloud-based database systems and the challenge of powering hyper-scale events such as the annual 11.11 Global Shopping Festival (Double 11), the Alibaba Cloud team started work on a robust, high-performance database designed specifically for the cloud. With as much as US$38.4 billion in sales clocked over a short span of just 24 hours and over US$1 billion of GMV in the first 68 seconds of its 11.11 Global Shopping Festival in 2019, this was by far the biggest test to validate the promised agility and scalability of any database system.

Put through its paces in real-world use cases such as 11.11 Global Shopping Festival, the world’s largest online shopping event, PolarDB’s ability to handle 544,000 orders per second and reach a new high of 1.292 billion total delivery orders is testament to its ability to support scenarios that demand extreme elasticity and scalability for enterprises in Asia-Pacific as organisations and businesses embrace digital transformation.

Created for the enterprise

For its raw performance and scalability, PolarDB isn’t just designed for massive e-commerce deployments, but also mission-critical workloads for enterprises. Built to be fully compatible with mainstream database systems, there are versions of PolarDB that are compatible with popular database systems such as MySQL, PostgreSQL and even compatible versions of Oracle.

Enterprises can hence migrate their applications onto PolarDB with ease, from financial applications to CRM and ERP systems. This can be done without completely revamping current application code or making extensive tweaks to the existing application stack. The result is a significantly easier pathway to move existing deployments over to PolarDB and the cloud.

A wide range of organizations are now using PolarDB and leveraging its capabilities across different verticals, ranging from manufacturing, public sector financial institutions, to global online businesses. For instance, PolarDB was recently used at scale to manage operations of PrestoMall, one of the largest e-commerce providers in Malaysia, enabling it to process millions of queries per second and support up to 100 terabyte of data.

PolarDB cloud database

Released publicly in September 2017, Alibaba Cloud’s PolarDB is a fully managed, cloud-native relational database system that promises significant benefits around performance, scalability, resilience, and cost-effectiveness. With support for low latency read replicas for scalability, this translates to an ability to scale up to millions of queries per second.

Below are some of the key capabilities of PolarDB.

  • High availability: One paradoxical aspect of a distributed database systems is their heightened propensity for failure – more nodes equate to more hardware for a higher probability of failures. To ensure that the system never fails, high availability support is built-in by default, preparing systems to resume normal performance quickly in case of any hardware failure. This has translated to an availability as high as 99.99999% with the use of three replicas within a PolarDB cluster.
  • Performance: To enable high-performance and scalability for the most demanding workloads, PolarDB operates as a cluster using a storage and computing-separated architecture. Each cluster consists of a primary instance that can grow to a maximum of 16 nodes with a database size of up to 100TB. As data copies are shared among multiple database nodes, storage cost is significantly reduced, while the distributed Polar file system means that database storage capacity can be automatically scaled up as necessary.
  • Cloud-native scalability: The standalone architecture of traditional database systems is hard-pressed to meet the increasing demands of cloud services. Built from scratch to meet the demands of cloud workloads, the cloud native PolarDB incorporates high scalability by default. Should higher requirements be required, its distributed “Shared Nothing” architecture means it can scale further with multiple PolarDB clusters.
  • Integrated design: In response to changing market demands, PolarDB enables its software techniques with its latest hardware advancements to offer world-class performance in data processing, storage and scalability. This ranges from NVM support for faster solid-state drives access, the ability to harness advanced GPUs and FPGAs, as well as Remote Direct Memory Access (RDMA) for high-throughput clusters.

Engineered for reliability

To ensure seamless operation, a host of intelligent capabilities offer self-optimization and self-recovery capabilities. This includes the high availability capabilities for automatic and rapid failover to operational nodes in the event of localized problems, far faster than the ability of even an alert and trained database administrator to notice, diagnose and manually mitigate.

At a recent 11.11 Global Shopping Festival, for instance, PolarDB was able to automatically detect a failing node and migrate the affected workloads to a healthy node while maintaining data consistency. Automatically performed within the span of 15 seconds from the onset of problems, this was possible due to the real-time replication of petabytes of data in the background.

In addition, PolarDB boasts an extensive array of technologies under its hood. It decouples storage and computation, leveraging RDMA over a fast network for additional performance and elasticity. Finally, it can achieve a dramatic reduction in storage overheads with a novel indexing structure and efficient space usage through next-generation storage optimization and code generation technologies.