People of a certain age will recall with nostalgia flicking through the Yellow Pages to find businesses to procure from. A mainstay of households, the hefty book was a trusted friend to turn to when your boiler packed in or your drains were blocked. But when the internet came along, and the revolutionary powers of search engines, the directory’s days seemed numbered.
Far from it – the directory format lives on today, albeit firmly adapted for the digital age. France’s version of the Yellow Pages, PagesJaunes, is one of the country’s most visited websites. The directory, which holds the details of nearly five million professional brands alongside access maps, historical commercial data and user reviews, is used by one in two French internet users.
Adapting to the times, however, has required a significant amount of transformation. PagesJaunes had to transition from a completely physical, printed directory, to a digital version without losing relevance. Yet still, its initial tech footprint would only take it so far.
“At the beginning, the [digital] solution was a monolithic one. Naturally, over time, that became difficult to maintain,” says Bertrand Riandière, technical lead architect at Solocal, the digital marketing firm which owns PagesJaunes. “A strategy shift occurred in 2013 when we moved to a microservices architecture, and questions arose about the existing MySQL databases…
“It’s crucial to respond very quickly to user requests,” he explains.
“Requests need to be answered in just a few milliseconds and must be relevant, particularly with the search engine. Solocal conducted studies to find and test databases available at the time. After considering several options, it turned out that MongoDB was the most advanced in meeting our specific needs, such as monitoring.”
Dawn of the cloud
The initial on-premise MongoDB setup was successful for its time, resulting in better availability, performance and easier maintenance. But as the cloud era arrived and the directory’s online users continued to grow, it became apparent that managing the infrastructure on-premise was becoming costly and prevented the site from evolving at speed.
(By this point it was running one of the world’s largest on-premises MongoDB deployments.)
In 2017, Solocal leaped and started to migrate its technical stack to the cloud. For PagesJaunes and Solocal, this transfer to IaaS was an opportunity to evaluate other database technologies. It decided after extensive diligence that MongoDB remained the best option.
The management and provisioning of the databases remained in-house for a couple of years. However, the need to reduce infrastructure costs further, and take full advantage of the automation, performance and security benefits of the cloud, meant moving to a fully managed database service was an inevitability. In 2019, Solocal committed to migrating its databases to a PaaS model using MongoDB Atlas, the fully managed global database service, on Google Cloud.
“The goal with the MongoDB Atlas deployment was to remove some modules we had to maintain internally, reduce costs and improve efficiency to speed things up,” says Riandière. “Now, indexing happens automatically, eliminating the need for an intermediate step to perform searches. Also, for asynchronous processing, we manage each step of an asynchronous task.
“Before, for our particular use case, we had a virtual machine running Elasticsearch, a Mongo-mirror to index all MongoDB updates into Elasticsearch, and some custom solutions for specific edge cases we had developed separately. But now, we no longer need additional clusters – everything is in the same place. In terms of time, there’s less maintenance required because it’s part of a managed service, which we already pay for. ”
More to come
Moving to MongoDB Atlas has paid dividends. Availability has increased to 99.995%, while taking advantage of automation and the managed services in Atlas has helped to dramatically slash database costs, which today account for just 4% of Solocal's IT budget, down from 30% previously.
Previously, a team of developers needed an entire week to transition to a new release but now a single operator can complete the process in just an hour. Backups and restores are also automated, which along with the auto-scaling option in Atlas, has given developers the time to focus on higher-value tasks like developing new features.
There is yet even more value to unlock.Solocal is transitioning more use cases from ElasticSearch to Atlas Search, consolidating its search capabilities under a single platform and simplifying operations. A recent example is system testing, one of the several validation phases before a module goes into production.
“Ensuring the correct data for automated testing involves a lot of criteria,” Riandière adds, “such as does the professional have their availability set? Are there sufficient data points? Do they have certifications, for example in construction or another field? Now we’ve switched to MongoSearch and as a result everything is now stored in MongoDB and we no longer need to maintain separate infrastructure. It’s dramatically simplified our operations.”
Published in partnership with MongoDB