As the keeper of the cloud-computing backbone behind Alibaba Group’s giant online marketplaces, AliCloud processes a massive amount of consumer transactions, web analytics, and logistics data each day. Indeed, at the height of this year’s 11.11 Global Shopping Festival, AliCloud was busy processing 140,000 transactions per second.
This practical demonstration of raw processing power came on the heels of a major technical achievement by AliCloud, as it recently set world records in two international benchmark tests as part of the Sort Benchmark series devised by the late computer scientist Jim Gray. Specifically, FuxiSort, AliCloud’s distributed computation framework, claimed first place in the GraySort and MinuteSort benchmarks, which measure how quickly a system can sort data. FuxiSort was able to sort 100 terabytes of data in just 6.3 minutes—a dramatic improvement over the previous record of 23.4 minutes set by Apache Spark in 2014.
Besting some of the world’s most advanced computational systems helps establish AliCloud’s cred as a competitor in the $210 billion global cloud services market. In its most recent quarterly results, Alibaba reported its cloud-computing subsidiary generated $102 million in revenue, up 128% compared with the same period in 2014. While Chinese customers are currently driving this growth, AliCloud has made no secret of its aspirations to compete worldwide with established Western cloud companies such as Amazon Web Services (AWS), Google Cloud Platform and Microsoft Azure. In 2015, the company opened two data centers in Silicon Valley and has plans to launch new ones in Japan, Dubai and Germany in the near future.
“The sorting competition basically showcased our core capabilities, much like the engine of a car,” says Tang Hong, Chief Architect of AliCloud. “But there are many other components that make a car a high-performing and desirable product to buy. We are trying to sell the whole car, not just the engine. So in that sense, the core computation capabilities are very fundamental, but there are still a lot of other pieces that make up AliCloud, like server virtualization, network virtualization, security, storage, data management. All of that is central to the whole cloud offering, and we’re heavily investing in those areas.”
While just part of the picture, the Sort Benchmark wins are validation of a long-running learning process at AliCloud. “This is really the result of at least five or six years of continuous development focusing on the performance aspects of the system,” says Tang. He explains that when data is processed in large volumes, there are always limitations imposed by the underlying hardware and bottlenecks in the network. But AliCloud has honed its systems to utilize both hardware and software resources efficiently to overcome those limits. Plus, this robust cloud-computing platform, which serves about 1.8 million customers, benefits from having thousands of parallel processors to enlist simultaneously. The benchmarks were set using a cluster of 3,377 servers, to be precise, which were employed from among “hundreds of thousands” of machines, Tang says.
The AliCloud team just needed to place a few calls to reserve the hardware. “We had the luxury of a fairly large testing facility to run this benchmark, and we mobilized and basically vacated the whole cluster to do so,” says Tang. “We called our tenants using the cloud, explaining that we wanted to run a benchmark over the weekend and asking if they could give us maybe a couple hours’ window.
“We could’ve run at an even bigger scale,” Tang says, “but we didn’t have a large enough cluster available.”
As Tang points out, sheer computational speed is not the only metric that matters to customers. The biggest benefit of AliCloud, he explains, is that any business can quickly build and deploy their own web portals, smartphone apps, and other software without going through the trouble of having to configure complex backend servers in data centers themselves. AliCloud features a comprehensive suite of built-in software that enables companies to run their cloud-based businesses straight out of the box.
With proliferation of mobile devices and the rise of the Internet of Things (IoT), ever more data needs to be accessible in the cloud, where users can access it from anywhere across a growing variety of devices. Customers also want to gain faster, more accurate and actionable insights without having to invest in computing hardware that will soon be outdated.
The GraySort benchmark suggests that AliCloud is on the right track when it comes to addressing customers’ needs, building its cloud with an “engine” that is already ahead of the curve. And while AliCloud has an uphill battle ahead of it to compete for customers internationally, it is clearly a leader on its home turf. A diverse range of businesses in China—including Sinopec, China Customs, BGI, and Intel—are currently using AliCloud.
Sinopec, Asia’s biggest oil refiner, uses AliCloud to store and analyze big data. Areas currently being explored by the collaboration with Sinopec include IoT, online payments, networked vehicles, e-commerce, and online-to-offline commerce (O2O). AliCloud is also helping China Customs to combat and prevent smuggling, while also optimizing the collection of import duties and taxes. Likewise, Intel Corporation and BGI have chosen AliCloud to support their work in the field of precision medicine and its applications, using big data analytics and cloud computing to expedite their work.
A survey by Netcraft earlier this year showed that thanks largely to AliCloud’s growth, Alibaba Group had become the 4th-largest web-hosting company in the world and the largest in China. By opening data centers at key locations around the globe to meet the demand and provide backup capacity to its existing infrastructure, AliCloud hopes to continue to gain speed.
