Rangespan provides an automated supply chain service that enables online retailers to massively expand product selection without risk—helping retailers get the right products from the right suppliers to the right customers via a hosted supply chain service. The company provides its customers with an order management system, supplier marketplace, customer protection system, and product data catalog in one end-to-end service. Rangespan was launched in 2011 in London, England and counts among its customers Argos, Tesco, ASDA, eBuyer, Maplin, Staples, and more than 200 other UK retailers and suppliers.
Holidays are peak season in the retail industry. To compete, retailers must be able to expand their catalogs by an order of magnitude—and supplies must be prepared for order volumes to increase by two orders of magnitude. For Rangespan, who serves both retailers and suppliers, the ability to turn on a dime is critical. The company needs to keep its inventory and price data fresh for more than 15 million products, and uses machine learning algorithms to keep up with the constant pace of consumer demand.
From the start, the company knew it would need an infrastructure that would enable it to expand beyond the U.K., throughout Europe and the U.S. To do that, Rangespan needed to manage and protect data across regions without making significant investments in physical infrastructure. It also required the ability to quickly experiment with technologies, services, and architectures.
Why Amazon Web Services
The company selected Amazon Web Services (AWS) as its platform so that the business could grow unencumbered by the costs and inflexibility of a traditional on-premises infrastructure. “The range of services offered by AWS meant we could focus on the product rather than the infrastructure,” says James Summerfield, Technical Director at Rangespan. “Using AWS paid off for us—within six months of starting the company, we launched with Argos, a top 10 retailer.”
Rangespan is built entirely on the AWS Cloud. The company uses Amazon Elastic MapReduce (Amazon EMR) to build and validate its product catalog, process the supply and demand signals to analyze and predict future sales. “We run 10-15 Amazon EMR jobs across our data sets on any given day,” Summerfield says. To parse and extract information about products (such as the name of the item, its color, or its dimensions), Rangespan uses Natural Language Processing (NLP). The company uses Machine Learning (ML) to automatically categorize products and predict potential future sales for each retailer, which helps Rangespan’s customers prioritize which products to adopt. “We use Amazon EMR for ad-hoc web crawling, parsing data from web-crawled content, and for applications that use NLP and ML techniques,” Summerfield says. “Combined with Amazon EC2 Spot Instances, it makes for a very cost-effective solution.” Spot Instances allow users to name their own price for Amazon EC2 computing capacity. You simply bid on spare Amazon EC2 instances and run them whenever your bid exceeds the current Spot Price, which varies in real time based on supply and demand.
The company uses Amazon Relational Database Service (Amazon RDS) to store customer information, order history, and price and inventory information, and stores additional data in Amazon Simple Storage Service (Amazon S3). Rangespan also uses dozens of Hadoop cluster nodes, a redundant three-host MongoDB configuration, and RDS-MySQL (includes Multi-Availability Zone redundancy).
Rangespan runs its website on AWS, as well. The company uses three Availability Zones and Amazon Elastic Compute Cloud (Amazon EC2) instances for its web servers and app servers. Amazon Route 53 is used for DNS, and AWS Identity and Access Management (IAM) is used to manage user access.
Amazon CloudFront provides a highly reliable and performance image content delivery network (CDN) for Rangespan customers. The company uses Amazon Elastic Block Store (Amazon EBS) with Provisioned IOPS to tune its MongoDB databases for maximum performance, and uses Amazon Simple Email Service (Amazon SES) for internal reports and partner notifications (for instance, when an order is shipped or a refund is made). Amazon Simple Queue Service (Amazon SQS) handles Rangespan’s asynchronous tasks (such as generating PDF packing slips and assigning orders) so that the company can maintain a highly responsive system for its users. See below for an illustration of Rangespan’s architecture.
Figure 1: Rangespan Architecture Diagram
By using AWS, Rangespan can support more than 5,000 transactions per second on its product and inventory data sets, and stores 400 million records in a 2 TB catalog. “Our suppliers benefit from new sales channels and our retailers are able to analytically select from a huge range and launch them easily, providing their customers with the products they want. AWS helps make that possible,” Summerfield says.
Using AWS has enabled Rangespan to focus on features that add value to the business. "Using AWS gives us peace of mind and saves us time that we can spend on innovation—it’s very valuable,” Summerfield says. “Year over year, we’ve experienced 400 percent growth in traffic and transactions—and using AWS helps us do that.”
The company has also reduced costs by using AWS. Not having to maintain an on-premises data center has saved the company anywhere from $800,000 to $1,000,000, Summerfield says, and the company saves 80% over on-demand instances by using Amazon EC2 Spot Instances.
By using AWS, the company has created a continuous deployment environment where each code push causes Rangespan’s full suite of tests to run in AWS. The company releases about 15 times a day. “We take advantage of spinning up new instances to test new ideas quickly,” Summerfield says.
Summerfield describes how Rangespan will use AWS products over the next 12 months: “I expect to expand and further automate our use of Amazon EMR, introduce new hosts for scalability and redundancy using Amazon EC2, and increase the capacity of our Amazon RDS. I may be scaling up our MongoDB instances for Amazon EC2 as well, adding a data warehouse using Amazon Redshift, and migrating some or all of the platform to a new geography. We’re also contemplating use of Amazon DynamoDB.”
"The reliability and scalability of AWS provide peace of mind, allowing us to focus on invention, not operations," Summerfield concludes.