bluebay2014 - Fotolia
Private clouds are commonly adopted to help enterprises leverage cloud technologies, while adhering to governance, risk and compliance requirements. Enterprises are beginning to learn that private cloud infrastructure can also be deployed onto bare-metal platforms to achieve better performance. This can be critical in industries such as finance, exchanges and transactions processing systems.
Paddy Power Betfair, the world's largest betting site, recently turned to private cloud deployment to future-proof its infrastructure. "We wanted to provide much faster deployment, and did not want to have to wait days or weeks for developers to get a new box or to set up a storage array," said Richard Haigh, head of delivery enablement at Paddy Power Betfair, based in Dublin, Ireland, at the Red Hat Summit.
The company runs the world's largest betting exchange, with about 1.7 million active users. Its existing private infrastructure manages about 135 million daily transactions, which leverage about 3.7 billion API calls per day. A private cloud deployment on top of OpenStack is already demonstrating better performance, compared with its traditional infrastructure. But the company decided to proceed slowly to work out some of the bugs.
Manage infrastructure as code
Betfair decided to look at the business need for next-generation business infrastructure about a year ago. There was a need for more scalability. Its existing infrastructure was good, but the company was growing, scaling higher and needed infrastructure that could scale up more easily. The company is a regulated business, and it needed compliance and security baked in.
There was also a need to work with continuous deployment to support software development and infrastructure that included compute, storage, networking and switching. "We wanted to provision all of this with the same mindset as managing code," Haigh said.
Until recently, this was not easy to do on OpenStack. It is a complex architecture, with components at different levels of maturation. For example, the networking model was limited. To address this challenge, Red Hat has been creating a variety of templates and use cases to help companies get started more easily. Red Hat also added CloudForms to provide unified operations and management of hybrid cloud infrastructure.
Bare metal was key
The exchange had a stringent low-latency requirement to keep customers happy. This created a need for virtual and bare-metal instances. Many of the exchange components require high-performance infrastructure. At the same time, Betfair wanted to use the same tool chain to provision onto bare-metal servers and other cloud infrastructure.
A key component of streamlining this infrastructure required implementing a tool chain for orchestration and continuous delivery. Betfair adopted Jenkins for execution, GitHub for source code management, Artifactory for the artifact management, Qualys for security and Ansible for provisioning. Both Arista and Nuage were adopted to automate the provisioning of virtual networks. Ironic was used to provision applications consistently onto bare-metal servers.
Considerations at every phase
Richard Haighhead of delivery enablement at Paddy Power Betfair
Making a transition to the cloud can be daunting. So, Betfair decided to break up the deployment into multiple phases. The first phase started with a proof of concept (PoC), with the goal of standing up a two-zone OpenStack prototype. Haigh said it was important to provision the PoC across zones to get a better understanding on how the apps behaved as they communicated across data centers.
Next, they began building the seeds of a pilot production infrastructure in the two data centers. This process made it easier to find problems in communicating between apps on the new infrastructure and legacy apps on the old infrastructure. The goal was to ensure this worked as well as the old private infrastructure.
The company is now in the third phase of private cloud deployment. A team of about 80 developers is working to move applications to cloud-based clusters. "Some of the applications are easy to shift, while others have architectural changes that have to be considered," Haigh said. In some cases, they have been pleasantly surprised by the performance improvements. Initial testing found the distributed file model is an order of magnitude faster than the 9-month-old physical system.
Betting on private cloud deployment
In many organizations, developers are being lured to the cloud by faster development and deployment capabilities. They don't want to have to worry about the infrastructure. If enterprises can deliver those services in a public cloudlike way, there is no need for developers to go around the private cloud infrastructure.
In the long run, Betfair is hoping to make it easier for its developers to take advantage of modern cloud platforms in the same manner. "We are trying to get something that is as easy as AWS [Amazon Web Services] for developers, but which we own and manage," Haigh said.
The worries of private cloud deployment
How cloud-native apps are changing private cloud
Guide to deploying private clouds