PreviousNext has worked closely with a large government agency on their public portal for the last few years. As one of the most visited Drupal sites in Australia with more than 7 million page views per month, the agency was seeking higher performance and cost efficiency from their Drupal hosting provider.
The government agency had originally engaged a prominent global Drupal hosting company, who while using Amazon Web Services for their underlying infrastructure, couldn't take advantage of contemporary cloud services such as managed applications and true autoscaling. This mean that any infrastructure upgrades in response to growing visitor traffic required manual intervention and its associated costs and time delays.
In early 2016 PreviousNext took over full responsibility for managing the client’s complex cloud services and used it as an opportunity to launch a next-generation cloud hosting platform we’re calling “Skipper”. Skipper leverages the best of contemporary cloud services on Amazon Web Services (AWS) or Microsoft Azure infrastructure, PreviousNext’s cloud infrastructure partners.
Managed Application Services
In order to provide a High Availability architecture, in Skipper we moved the MySQL database to a hosted Relational Database Service. Setting this up in Multi-Availability Zone mode meant that not only are there no issues when a database instance goes down, but ongoing maintenance and updates are handled automatically. This dramatically reduces the manual maintenance required. We also took advantage of using Simple Storage Service for shared files, which avoids the complexity of managing ever-growing disks and storage constraints.
A key requirement of the agency was to host multiple sites on a single High Availability platform. The platform now hosts two large portals alongside a number of testing and staging environments. The use of Kubernetes allows Skipper to be extremely efficient in compute resource utilisation, so we’re only paying for infrastructure we use. This is in stark contrast to the huge fees associated with sizing infrastructure for a worst-case traffic scenario, as was the case with their legacy hosting provider.
One of the key benefits of the architecture is its ability to truly autoscale based on visitor demand without manual intervention. If site traffic increases, more Docker containers are automatically added to increase capacity. If the number of containers reaches a threshold, then more worker hosts are deployed. The result is a highly flexible and fault-tolerant platform that can scale to withstand spikes in traffic and scale in unused compute resources to save money.
PreviousNext have been actively transitioning a number of our large government and enterprise clients away from their legacy hosting providers to new Skipper cloud platforms over the past year, resulting in significant gains in performance, flexibility and cost savings.
We’re now seeing demand for Skipper from large Drupal users that PreviousNext hasn’t had a prior relationship with. These clients are seeking all of the benefits Skipper provides, but want to maintain their site using their existing internal team or incumbent agency. In response to this, we’re in the process of developing more client friendly interfaces to provide access to information on the cluster, how many containers are deployed, notifications when scaling events occur and expose our Continuous Integration
While Skipper is not a mass-market, self-service offering, we do see it having a bright future with clients that are seeking an alternative to fully managed enterprise Drupal hosting services and want to take advantages of all the latest advances in cloud services.