GCP faster and more powerful, says Telegraph
The Telegraph has gone all-in on Google Cloud Platform (GCP), ditching its hybrid cloud structure – which previously combined an electic mix of infrastructure including an on-premises data centre and elastic cloud provision from AWS – for operations based entirely on GCP, from advertising campaign analytics to image optimisation.
The 160-year-old broadsheet publication has a print circulation of approximately 360,000 daily and claims to reach 23.1 million people a month across all platforms. It plans to use cloud-based machine learning capabilities to personalise the MyTelegraph app, predict the demand for physical newspapers, and classify content for journalists – including Telegraph images – to improve “discoverability” and upload speed.
Lucian Craciun, Head of Engineering & Technology Platforms at The Telegraph, told Computer Business Review in a call: “We used to parse logs out of AWS with some earmarked clusters. It took hours. A few years ago we started using (GCP’s cloud data warehouse for analytics) BigQuery; it was really quick and really cheap; magic.”
(A Google case study says it now processes daily advertising campaign performance data in eight minutes, compared to The Telegraph’s previous 2.5 hours.)
Craciun added: “One of the other issues we were facing was notifications for readers on the MyTelegraph app. One of the constraints was that readers should get a real-time notification when a story about a subject or by a reporter they follow is live. When you scale that to millions of devices it starts to get complicated; you have to send millions of notifications at the same time. Firebase and Firestore (GCP’s web application development platform and scalable database for mobile, web, and server development) solved that problem for us, so we started looking at some other Google products.”
As he details in a blog for GCP, The Telegraph has set a target of ten million registered users for its My Telegraph in the next few years.
“Obviously, this service is heavily dependent on APIs. Before My Telegraph, readers had access to a curated feed of editorial articles and they had to use the website’s search function if they wanted to find content related to specific topics or by individual journalists.”
The paper now has a range of personalisation options that correspond to individual APIs: “What is new for us is having a single gateway in front of all of our APIs. Previously, we used an API gateway for some, a straight load balancer for others, and a content delivery network for still others. As a result, all our APIs didn’t have a structured way of being exposed to the public or to other teams. What Apigee introduced was much-needed structure in the form of a single gateway for all our APIs. We now can easily find what we need when we need it, which is a tremendous help as we look to develop and deliver new services”
The Telegraph was already using Docker, with a scheduler to manager containers, but everything was on AWS and, as Craciun puts it, “it’s not easy to get engineers who know it. The industry standard is Kubernetes so it made sense to look at Google offerings for Kubernetes. GKE (Google Kubernetes Engine) was there and easy to set up…”
Google added: “The Telegraph continually loads clickstream logs, advertising logs, and a range of other first- and third-party data sources into BigQuery for a deep and connected view of all behavioral data for analytics. By analysing data from many sources, the publishing house gains a deep understanding of what interests its readers, which informs editorial, marketing, and advertising decisions.”