MWM Uses Google Cloud to Coordinate Its Growth Worldwide

From PaaS to IaaS

Since 2013, MWM has used the App Engine virtual servers to port the code for its apps, which are developed in Python. "This means we can focus on the quality of apps without being hampered by deployment, availability, updating, and maintenance issues," says Damien Delépine, Head of Cloud, Web & Big Data at MWM.

The company found that, on average, App Engine can process 300 to 500 requests per second, with peaks of up to 800 simultaneous connections. In 2016, MWM went one step further by changing a third of its services from a platform (PaaS) to a managed infrastructure (IaaS) and container (CaaS) method.

Compute Engine was then complemented by Google Kubernetes Engine . "The architecture of Google's container system separates the different parts of our apps while retaining a high-performance level,” says Damien. “In addition, it has been really easy for us to deploy and test new technologies, such as search engines, user configuration storage services, or even databases replicated on dozens of virtual servers."

Using software containers to separate roles isolates the process of each task within apps. This optimizes resources as the hardware configurations are adapted to each task's profile, according to what they consume the most: computing units, RAM, or storage. MWM currently uses around 40 servers, which are divided into four clusters that are optimized for containers.

Porting code into Google Cloud is facilitated by Container Registry , which carries out runtime tests before building the apps and sending them into production. Persistent Disk retains important data and "reinjects" this into the production servers at any time.

Finally, "the Google Cloud operations suite monitoring service has quickly become an absolutely crucial decision-making tool,” says Damien. “For example, it revealed that our music-video player generates large outgoing data streams from our servers. We upgraded our apps and now compress this data. Stackdriver therefore contributes to our technical development planning and helps us control our costs."

One Vision

Since 2015, MWM has also used Google Cloud's analysis services to gain an overview of its apps and its customers. By analyzing hundreds of gigabytes of operating data each day, BigQuery highlights potential problems with specific features, regardless of the servers, databases, and mobile platforms (iOS, Android and Windows) concerned. Thanks to this comprehensive overview, the company has been able to improve user retention and monetization of its services.

In terms of marketing, BigQuery supplies community features: shared music lists, charts by genre or country, etc. "By linking BigQuery to the App Engine memory cache service, we are able to offer our users unbeatable service quality," says Damien.

Relational and Non-Relational Databases

MWM uses Cloud SQL to store data from its app storefronts, user interfaces, its corporate website, and even its tutorials and FAQs. Syncing data between apps and data backup is performed using Cloud Storage , while Datastore unifies user accounts across all MWM applications.

Cloud Spanner is also used: "This is a standard but globally distributed relational database, which gives us the best of both worlds: SQL and non-SQL,” says Damien. “We followed the beta version closely, and will soon be using Spanner in production for one of our services."

"All in all, Google's product range is completely flexible: it gives us free rein over the deployment, coordination, and monitoring of our apps. It helps us get a better understanding of our customers and plan development of the features they want. It's an essential tool to help companies like ours in the mobile industry grow quickly," concludes Damien.