Powerspace: Creating an ultra-stable stack for a more productive team

About Powerspace

Based in Paris, Powerspace specialises in digital advertising services, serving innovative non-intrusive ads for French ad publishers and agencies.

Industries: Other
Location: France

Powerspace uses Google Cloud Pub/Sub, Dataflow and BigQuery alongside Google Kubernetes Engine to create an ultra-stable stack, delivering low latency Big Data analytics with minimal maintenance.

Google Cloud Results

  • Helps ensure high availability by eliminating system dropouts
  • Frees the coding team from infrastructure maintenance tasks to concentrate on their core mission
  • Saves a projected 25% on server costs thanks to transparent billing and efficiency savings

Projected infrastructure savings of up to 25%

By offering innovative non-intrusive advertising services to digital advertising publishers and agencies, Powerspace has become a leader within its market in France. It specialises in real-time advertising within email newsletters, serving about 1 billion ads per month across all devices. Since its foundation in 2008, Powerspace has developed all its advertising technology in-house.

In order to run its platforms and deliver high-quality ad recommendations, its system needs to process several terabytes of complex data hourly with very low latencies. Its services also need to be highly available, so that no ad impression is lost, and clients can access the platforms at any time.

“It is relatively hard to deliver real-time dynamic advertising within email newsletters that is personalised for every single recipient. Our key technical priorities are extremely high availability, which we need for our advertising solutions, and the capacity to process and feed back very large quantities of data in order to constantly improve the quality of our advertising offer.”

Cédric Sadai, Co-founder, Powerspace

Powerspace’s in-house approach included relying on a physical hardware center at its Paris-Opera headquarters, which was not reliable enough, time consuming, and insufficient for its needs. Google Cloud Platform was the perfect fit to solve these problems.

“It is relatively hard to deliver real-time dynamic advertising within email newsletters that is personalised for every single recipient,” says Cédric Sadai, co-founder at Powerspace. “Our key technical priorities are extremely high availability, which we need for our advertising solutions, and the capacity to process and feed back very large quantities of data in order to constantly improve the quality of the advertising offer.”

“We managed to move several billion logs a month via Cloud Pub/Sub and we never lost a single message whereas previously, we lost at least a few messages every month. For us, that’s the real value—stability.”

Cédric Sadai, Co-founder, Powerspace

Migrating with zero data lost

Digital advertising depends on getting the most relevant advertisement to the end-user through auctions that take place continuously. In order to serve its publishers and agencies, Powerspace needs a system that delivers processing power day and night and that is also very reliable. Previously, its physical data center was prone to breakages and electrical faults and as a result, its services suffered unexpected dropouts, causing anxiety for the Powespace team. It looked for a solution that could deliver high performance as well as increased stability.

To do that, Powerspace switched to Google Cloud Platform. Using Cloud Pub/Sub, it created a parallel flow of data in order move its data warehouse step-by-step. According to Cédric, “Cloud Pub/Sub was the technology that allowed us to launch the whole migration project. We had a stack in production, serving clients and processing a billion impressions a month. We needed to transport this into the Cloud without a single second of downtime.”

Once Cloud Pub/Sub was handling the flow of logs, Powerspace began to push data through Cloud Dataflow to Google BigQuery. “Once we were certain that the data was definitely equivalent and we had the stability we needed, we migrated that part onto the Cloud,” says Cédric. “Then we proceeded the same way with the rest of our stack.” In the next phase of migration, more strategic applications serving data in real time, as well as its middleware and extranet were all moved to Google Kubernetes Engine. Throughout the whole process, the stack remained fully functional.

“We managed to move several billion logs a month via Cloud Pub/Sub, and we never lost a single message,” explains Cédric. “Whereas previously, we lost at least a few messages every month. We have never had a single instance of downtime on Google BigQuery. For us, that’s the real value—stability.”

A more confident team

Migrating Powerspace’s stack to Google Cloud Platform has had a positive impact on team morale. Before, unexpected problems arising with physical hardware caused the team unnecessary stress. “With an on-premises system, the team spent an enormous amount of time managing problems. There was a lot of DevOps and fine-tuning to do,” says Cédric. “That really put pressure on the team.”

Now, they have greater confidence in their system as well as more time to focus on their jobs. It has also made welcoming new team members easier. “The migration has liberated the coders to actually work on their core mission, and the onboarding of new arrivals has much become simpler,” says Cédric. The team has more confidence in the product and the technology that is behind it.”

“Our strategic objective was to accelerate our go-to-market strategy and to stop wasting time repairing software and hardware faults. Instead, we can concentrate on our unique value advantage and deliver that to our customers, and expand our geographical reach. The migration to Google Cloud Platform is going to be a backbone for our future strategic deployment.”

Cédric Sadai, Co-founder, Powerspace

Delivering greater value

As a result of moving to Google Cloud Platform, the Powerspace stack is more stable while maintaining the low latencies needed. “We now have a system that is very resilient, with high availability, and that is very fast: we have a latency of just 6 milliseconds on Cloud Bigtable,” says Cédric. “We no longer feel like we are at the limits of our system, and the team has gained in productivity and well-being.”

Thanks to transparent and flexible billing, the projected financial benefits are also considerable. “We are likely to make a saving of around 20, maybe even 25% compared to the stack we had previously,” says Cédric. “A large-scale data warehouse would require a robust cluster of 20, 50, 100 physical machines to monitor and manage, whereas now, we just redirect our logs to BigQuery. And in the end, we only pay for the traffic that passes.”

Now, Powerspace plans to launch a SaaS version of its services and expand its market beyond France, as well as explore Machine Learning with TensorFlow, and Cloud AutoML to enhance its models for click predictions. “Our strategic objective was to accelerate our go-to-market strategy and to stop wasting time repairing software and hardware faults,” says Cédric. “Instead, we want to concentrate on our unique value advantage, deliver that to our customers, and expand our geographical reach. The migration to Google Cloud Platform is going to be a backbone for our future strategic deployment.”

About Powerspace

Based in Paris, Powerspace specialises in digital advertising services, serving innovative non-intrusive ads for French ad publishers and agencies.

Industries: Other
Location: France