How the next generation of databases are fuelling innovation
By Tan Jee Yee April 20, 2020
- During the last 24-hour 11.11 Shopping Festival, PolarDB dealt over 87 million requests per second
- The smarter alternative is to move to cloud-native databases, which can be scaled effortlessly
Kuala Lumpur-based PrestoMall (formerly 11street Malaysia) is an undeniable success. From 2015 to 2018, Malaysia’s largest home-grown online marketplace registered a massive growth of 256% in revenue.
The next phase of growth is inevitable and imminent, but herein lies a problem. PrestoMall’s previous infrastructure was not designed to meet the requirements of an elastic growth. What they essentially needed is a solution that not only offers elasticity and scalability, but also maintains cost efficiency.
They found the answer in Alibaba Cloud’s PolarDB, the cloud-native database that is designed for business-critical database applications, especially those that require fast performance, high concurrency and automatic scaling.
Going with PolarDB feels natural for PrestoMall. During last year’s 24-hour 11.11 Global Shopping Festival, Alibaba Cloud powered a whopping US$1 billion of GMV (gross merchandise volume) in a mere 68 seconds – during which PolarDB dealt over 87 million requests per second and supported the analytics that was generated by customer activity in real time.
Thanks to PolarDB, PrestoMall’s current infrastructure is capable of processing millions of SQL queries per second and can support up to 100TB of data storage. At the same time, PolarDB delivers better security, reliability and availability than traditional commercial databases. The company claims that over 40% cost reduction has been achievd after migrating their data and transaction workload to PolarDB.
Database as a cornerstone
Behind every technology innovation today are databases. For one, databases are everywhere, and is the foundation of the digital economy. Every major industry, from banks to securities to e-commerce, e-government, mobile payments, media services, online entertainment and education are all powered and managed by databases.
PrestoMall’s story is just one of many businesses that is rocket-fuelled by efficient databases. Some of the world’s greatest innovations and digital services today have strong database backbones. For example, you certainly can’t expect a fintech company with tens of millions registered users, to function without a database that is fast and efficient.
Yet, if PrestoMall’s case is anything to learn from, businesses and services undergoing digital transformation have to start thinking deeply about the databases hosting their entire operations.
For the most part, self-built, on-premise databases may serve its purpose. There are, however, limitations to this. “With self-built databases, there are various aspects to care for including the design and implementation of high availability, data storage and persistency, data security and more,” says Li Feifei, VP of Alibaba Group, President and Senior Fellow of Database, Alibaba Cloud Intelligence. “Then there’s the implementation of master-slave replication and RAID. While all of these look simple, they require lots of man-hours spent to achieve a high-quality database.”
What’s more, database servers can be costly. At the same time, users must purchase their own license if they are using commercial database products. “And the worse thing is the low ratio of resource utilisation such as CPU, memory and disk storage. It’s a tough task to manually achieve or build elasticity for your self-built database in order to save money.
“You either spend resources on accommodating peak traffic requirements, or handle the scaling with extra human operational costs,” Li adds.
The smarter, more cost-effective alternative, not to mention more efficient, is to move to cloud-native databases. For one, cloud databases can be scaled effortlessly.
“Traditional databases typically use a single-node architecture, whereas cloud-native databases usually use a shared storage architecture. This architecture separates storage from computing to enable the fast scale-out of computing nodes. Customers can use this shared storage database to complete a non-intrusive data migration without any change to the original business logic,” Li says.
“Cloud databases also help reduce the cost of data migration. Traditional solutions typically plan and resolve the read and write conflicts by using the OLTP system to process transactions and the OLAP system to analyse huge volumes of data. In the cloud native era, Alibaba Cloud can minimise the cost of data migration by taking advantage of the technical benefits delivered by new hardware devices,” Li explains.
This can be done by integrating transaction process and data analytics features in one database engine so that the two needs can be addressed seamlessly by one system.
Performance and scalability
Even so, not all cloud-native databases are built the same. Businesses handling massive loads like PrestoMall will need to look at features that scale and perform well during critical moments, for example, peak time usage.
PolarDB would be an example. Without getting into too much technical specifics, PolarDB integrates a collection of innovative technologies and takes advantage of the latest technologies of IT hardware including high-speed network and storage devices.
What’s truly unique here is that PolarDB separates compute and storage in distributed manner leveraging the underlying Alibaba Cloud infrastructure. “It provides efficient shared distributed storage over an RDMA network. The shared storage technology establishes a one-write-multiple-read model among multiple computing nodes,” Li says.
“In addition, the shared storage technology also allows for the rapid scaling of storage nodes based on the actual needs of customers.”
They also just pay for an appropriate amount of PolarDB resources according to their own application scenarios, as well as the peak and trough data volumes of their business. “This billing model significantly improves the efficiency and reduces the cost to save money,” Li explains.
NewSQL is another fascinating feature. Arriving around a decade ago, the system allows handling of massive amounts of data from decoupled resources in the cloud. “A company could scale from 100 nodes to 1,000 nodes in seconds, like e-commerce companies during a sale where traffic spikes. Alibaba has a partnership with MongoDB, who offered NewSQL technologies,” Li says.
“NewSQL is not just about scalability. It also gives you guarantees of consistency like a relational database. It has the best distributed and cloud-native architectures. We also have a hybrid database management system where we can run databases instances either running on-premises systems or in the public cloud,” he adds.
In addition, Alibaba Cloud database provides a wide range of purpose-built databases to meet various business application scenarios. “Alibaba Cloud enhanced Redis brings the strong data cache engine, which has been applied to support Alibaba e-commerce ecosystem more than 10 years, to the public cloud,” says Li.
This is part of Alibaba Cloud's vast portfolio of DBMS services – the largest among other cloud service providers in the Magic Quadrant, after all, and mixes commercial and open source offerings besides its own internally-developed or extended products. You get multiple choices for some types, too, such as PostgreSQL and MySQL for RDBMS. Alibaba Cloud also contributes to Apache HBase, the open-source non-relational distributed database.
It’s not to say that personal databases aren’t efficient enough to handle certain business processes. But for organisations looking to drive innovation and ensure that they stay on the competitive edge digitally should definitely consider embracing the next generation of databases.