Shazam turns to Google for GPUs in the cloud
Google offers bigger encryption menu for Cloud customers
The popular mobile app Shazam is moving to the Google Cloud Platform, it announced Thursday, after about five years of running its own bespoke non-elastic GPU infrastructure.
Shazam lets users identify a song they're listening to by matching it to the songs in its database. Launched in 2008, Shazam is used more than 20 million times a day.
The company has been using GPUs for its recognition services since 2012, starting with the Nvidia Tesla M2090. It now uses the K80. Google added beta support for Nvidia Tesla K80 GPUs on the Cloud Platform in February.
"We have traditionally used bare metal servers because GPUs in the cloud have not been available, and when they have, they were far too expensive and not performant enough for our needs," wrote Ben Belchak, Shazam's head of site reliability engineering, in a blog post. "Only recently have the economics of GPUs in the cloud really made sense for our business."
In a separate blog post for Google, Belchak said, "Other cloud GPUs weren't fast enough for our needs, but the Google Cloud Platform tooling ecosystem, its pricing, and Google's reputation convinced us."
Moving to the cloud should help Shazam run its service dramatically more efficiently. Previously, Shazam had to provision enough servers to meet peak demand and run at capacity around the clock. This was necessary since the app is subject to huge spikes in demand, which are at times predictable (such as during the Super Bowl) and at times unpredictable (if, for instance, an obscure song is featured in a popular commercial).
With rapid autoscaling on Google Compute Engine, Shazam can provision for 50 percent of maximum demand instead of 100 percent. So far, the company has migrated about one-third of its infrastructure into Google Cloud.
Belchak noted other benefits of moving to Google Cloud, such as the ability to quickly replace failing nodes, as well as the ability to update Shazam's audio signature database more frequently.