KubeCon Keynote Highlights Condé Naste International's Journey to a Distributed Kubernetes Platform
June 6, 2019
At KubeCon + Cloud Native Conference Europe 2019 last month, one of the most interesting keynotes came from Katie Gamanji, Cloud Platform Engineer at Condé Naste International. With well-known brands like Wired, Vogue, GQ and Vanity Fair under its umbrella, the prestigious print and digital publishing company operates in over 12 markets, including the U.S., Europe, China and Russia. Its 62 websites generate over 300 million unique users each month with 1.5 billion monthly page views.
The company previously allowed its different global territories to operate independently, which had led to a fragmented technical strategy across the company’s international markets. Gamanji joined the company a year ago to be part of a far-reaching project focused on the creation of a centralized platform intended to “embrace cloud native principles” and “further emancipate our international teams”.
In her talk, Gamanji summarized her team’s journey towards deploying a unified platform globally. Her talk particularly resonated with our team because of the many parallels between their architectural decisions and how we’ve designed the Section platform to help development teams achieve similar objectives.
The Condé Naste International Journey
Condé Naste International have been on a challenging journey towards delivering a multi-cluster distributed Kubernetes platform with a centralized management mechanism and self-service CI/CD process on a global basis. In her keynote, Gamanji covered the problems they were previously facing from a fragmented technological landscape, how they focused on building a solution for this and the resulting tech stack, in addition to challenges they continue to face.
A Fragmented Technological Landscape
Initially, the global market teams at Condé Naste acted like independent business units. From an engineering point of view, this meant that each market had an individualized tech stack with few shared components. It also led to fragmented design and visuals, which compromised the unified experience the publishing company wanted to be able to offer their worldwide customers.
China and Russia represent 17% of Condé Naste’s total digital readership and present unique content delivery challenges, spotlighting the need for the company to adopt a cloud agnostic infrastructure, and the benefits of deploying a self-hosted Kubernetes solution.
Issues around origin latency was another reason why Condé Naste began its journey towards global distribution of its servers. “What we aim for”, Gamanji said, is “market proximity and the highest user experience for our customers.” As such, it was a business decision to replicate its clusters in five different regions, making a total of nine worldwide to avoid latency for the end user.
A Unified, Cloud Native Platform
Two years ago, Condé Naste International began its journey towards building a unified platform across its multiple regions, embracing cloud native principles throughout. As with the Section platform, Kubernetes underpins the entire infrastructure. A Tectonic installer is used for their cluster deployment, which allows them to operate a self-hosted control plane for Kubernetes in addition to being able to plug it into numerous cloud providers to consume compute and networking resources. Similarly to Section, this allows Condé Naste International to avoid vendor lock-in while still deploying self-hosted multi-master, multi-node 100% komposers on Kubernetes.
Condé Naste International currently operates more than 100 instances in AWS, its cloud provider. Gamanji explained that they make use of auto scaling groups to “make sure that we scale up and down based on interaction with our customers with our brands”. Their infrastructure is deployed using Terraform. Like Section, they use a wide range of open source tools to provide authentication, log-in, monitoring and other tasks; one such example is Helm, their de facto deployment package manager to Kubernetes.
The new platform has been live since October of last year when Condé Naste International first launched GQ France. Since then, another ten websites have been launched with the overall goal being to migrate 34 websites.
Similarly to our goal at Section, Condé Naste International’s engineering team was focused on becoming a developer-first company, which empowers its local and international teams through the new unified platform. Gamanji discussed how having a self-service continuous integration/continuous delivery (CI/CD) process was an important step towards achieving this. We have likewise found that CI/CD pipelines are useful to help manage environments that have risk attached by breaking changes into small, verifiable units. This reduces the possibility of a major software flaw being implemented to an entire system, and helps to encourage a more responsive, quicker-time-to-market across the entire software delivery cycle, giving developers the flexibility to quickly, yet safely test out their code in production.
Continued Challenges for Condé Naste International
Gamanji finished her talk by highlighting some of the challenges they continue to face. These include service continuity; when migrating services to a new platform, the engineers need to maintain availability of content with minimal or zero disruption to users. Condé Naste International’s team have found that the best way to do this is to use Varnish configuration files, allowing them to implement changes to the cache with immediate effect. Gamanji also admitted that upgrades are a continued “pain point”. This is the result of their having a self-hosted solution for Kubernetes, she explained, which doesn’t allow them to do upgrades in an automated, sophisticated way. In particular, Gamanji mentioned the challenges of deploying infrastructure into China and Russia. Looking ahead, her team wants to introduce tracing, observability and service meshes into their clusters, an objective that the Section platform team is actively focused on.
She concluded by attributing her team’s success thus far in creating a cluster with a centralized management mechanism, which is globally distributed, to their “solid, well-defined CI/CD process that further empowers our local and international teams and their product”.
How Section Helps Development Teams Achieve Similar Outcomes
For those companies that may not have the internal resources to build a similar solution or indeed want to devote the time and finances towards it, Section’s Edge Compute Platform provides approachable tooling to manage a distributed microservices architecture.
The Section platform allows developers to focus on application development while we concentrate on ensuring that your workloads run in the most sensible locations to optimize performance, security, and scalability objectives. Browse our case studies to see how multiple industries, from testing to gaming to sales, are working with Section to gain control over their workflows and make edge programming a reality.
The full presentation can be viewed here: https://www.youtube.com/watch?v=D7pbISekc8g&feature=youtu.be
Presentation slide deck: https://static.sched.com/hosted_files/kccnceu19/c0/Katie%20Gamanji%20May%2022.pdf