The problem
- Need to operate a website with a wider variety of peaks and troughs
- Requirement to handle databases
- Ability to offer recommendations
- Easy expansion path
The solution
- Amazon Web Services offered price and flexibility
- Easy upgrade route
- Full redundancy
Running a flash sale sports clothing and equipment company online means that peaks in demand are an almost everyday occurrence. SportPursuit is a company that would struggle to exist without the cloud, it would simply not be able to cope without an infrastructure that could handle peaks and do so at a relatively reasonable cost.
The firm was started in 2011 and according to Steve Blake, chief technology officer of the firm, being a start-up means you never quite know how long you are going to be around.
“When you are starting a business, you want to do everything as cheaply as possible,” he says but adds that while this was the initial reason using the cloud, the biggest reason now is “because of our fluctuating traffic”.
“Also, with retail in the UK and the rest of the world, you have days such as the lead up to Christmas. Your traffic is just so much higher than during the first part of the year. So having an infrastructure where you can spin instances up and shut them down based on your demand is pretty key to cost saving, but also just providing users with an experience that every user thinks they're the only person on the website,” adds Blake.
Buying in infrastructure and finding a partner
In the run up to starting the website, Clapham-based SportPursuit evaluated a couple of hosting providers based in the UK: Rackspace and Dedipower. “But at the time, AWS was probably the leader when it came to cloud computing and [being] really affordable. That was the main driver when we initially chose [it].”
He says that at the time AWS was the only one that provided both flexibility and affordability. “Having a cheap hosting provider but also the lack of long term contract is quite attractive from AWS.”
Blake says the initial setup was “pretty basic” - just a web server and a database server.
“Then we’re running Magento, which is an open source e-commerce platform,” he says. “It was about installing the platform on the servers, running Ubuntu operating systems on the instances.”
His team spun up instances from AWS images and then installed the packages they needed. “And then from that, it has evolved and now we have got multiple web servers running.”
SportPursuit uses Amazon Relational Database Service to store product, order, and customer data, and Amazon Simple Storage Service for backup and storage of around 1TB of data. ElastiCache is used for in-memory caching and Elastic Load Balancing provides load balancing.
“We are using RDS database offering because it is MySQL-as-a-service,” says Blake. “It is completely hosted by Amazon, you just upload your database and they handle all the replication across servers. It is really easy to do back ups; it’s really easy to spin up new slaves if you need them.” He adds that this allows them to have a master and slave database for redundancy.
“We are also using Rediff and Memcache. We are using those out of the Elasticache service, which is just Memcache and Rediff as-a-service. Elasticache offers those two technologies as a service.”
The company manages its AWS cloud infrastructure using AWS CloudFormation, using it to create and manage a collection of related AWS resources. It also uses Amazon CloudFront to serve static resources around the world.
“We are using the virtual private cloud. Most of our instances are inside the virtual private cloud. From a security point of view it makes a lot of sense because the instances aren't publicly accessible,” he says.
Blake says that the also used Amazon’s Simple Mail Services in the past, but have now moved to dedicated email service provider for sending out emails.
Benefits
He says that having the cloud infrastructure in place has allowed the business to grow and that it is enabling them to hit business targets. “From month to month, sometimes we'll be in-line, sometimes we'll be slightly over, other months we'll be slightly under. Generally we are meeting our investor's expectations.” He adds that the firm has managed availability of its services for customers of more than 99 per cent.
He also says that by running its infrastructure in the cloud it has saved lots by doing so rather than on a traditional datacentre. He didn’t say just how much those savings were but if it had to use traditional infrastructure, the firm would have to run ten instances all day every day, instead of just having them during peak periods.
The cloud has also cut down on the time to provision. Blake says if he were using an in-house datacentre, provisioning would take three to five days whereas an instance with AWS takes only three to five minutes.
He says the decision to go with AWS was a good one as it has allowed them to grow from the one database server and one web server to around thirty instances running at any one time across its various live staging and UAT (User Acceptance Testing) environments.
Next steps
Blake says that within the next six to twelve months the business will concentrate on two things: expansion and a new recommendation engine.
“We have a user base in Australia and we are looking at going into Europe,” he says. “So along with that comes other currencies, translations, etc., etc. It's not just like the flick of a switch to get a German website up and running but that will be the first thing.”
The other focus of SportPursuit is the creation of a recommendation engine to spur more growth and greater revenues from customers.
“At the moment we do give a bit of personalisation based on people's gender and what sport we think they are interested in. But we'd like to give recommendations to each and every user based on their browsing behaviour, their shopping behaviour, etc.”
He says the goal would be to have this working along the lines of the classic Amazon 'here are some recommendations for you'. This could involve the firm looking to big data tools from AWS as well.
“Potentially Redshift but initially we are going to be using the Elastic Map Reduce service,” he says. “A great recommendation engine will take time.”
