One of the glib cloud concepts is scale. One app can reach 10 people or a million, or so marketing insists. Sometimes, though, reality matches the hype — and mapping the surface of the planet certainly qualifies as computing at scale. Mapping the surface of the planet with high-resolution VR images is another notch up. Yet one gateway to the most famous cloud service that does this, Google Maps Street View, is actually run by a 25-person company, Panoskin. And it all runs on MongoDB Atlas, the fully managed, global cloud database.
Tom Chomiak, CTO and co-founder of Panoskin, explains that scale works both ways. “We work with some of the biggest enterprise customers and agencies in over 100 countries,” he says. “But we also support individual photographers that want to build a virtual tour.” This side of the business has around 15,000 customers creating around 12,000 tours a month.
“If you’re a professional photographer, you can upload your work. We partner with Google Maps Street View, GoPro, and we can do what Google Street View scanning cars cannot — go inside buildings.” Strap a GoPro to a backpack and walk through the place you want to capture — the VR environment will be built automatically and integrated into Street View. “We’ve done entire countries with the lightest of resources. One man is on a mission to capture all of Lebanon. But if you want to map a sports stadium or a hotel, you can do that too, just by walking around.”
Chomiak estimates there are some 60 million images in the database. Taken together, they probably cover a distance equivalent to half-way to the moon.
The system is built on hybrid cloud, Google Cloud Platform, and Linode. And after four years using self-hosted MongoDB, the company is in the middle of transitioning to Atlas. “We’ve been staging for six or seven months on Atlas on GCP,” says Chomiak. “You have to be extremely sure you’ve got the right solution when you have limited developer capital yet you’re working with such big customers. We absolutely couldn’t do anything without MongoDB, and Atlas is proving a very good fit too.”
He says, although Panoskin is still quite new to Atlas, the migration path is logical. “We initially got started with cloud tools, monitoring, and backup tools. We found that, once the developers started using Atlas, they wanted to move everything into it. Reporting, running updates — the interface is incredibly intuitive.”
Without needing to worry about database administration, he says, the company is free to “build quickly and deliver on enterprise scale”. The hybrid cloud environment has delayed moving the production environment, but not by too much; it’s just taking a little time to migrate application servers to GCP. MongoDB is the only database Panoskin uses, with the exception of Redis caching that handles reads at the edge.
So far, says Chomiak, the company has self-hosted MongoDB in a single data center rather than in a geographically distributed system, the different regions of GCP help with this strategy. “This drew us to Atlas, as application servers can be load balanced in different data centers, which helps with local latency.” Latency is one of the key aspects of providing a good experience, he says — that makes it important to have intelligent caching and to keep data close to those who are most likely to use it.
We have a cloud-first approach now. It looks cheaper to self-host, but that doesn’t take into account the DevOps processes we get on the back end in Atlas. We don’t get concerned about maintenance and we don’t hire DBAs. For a small company, it’s a wise decision.Tom Chomiak, Chief Technology Officer, Panoskin
“We’re a 25-person start-up from Chicago, yet we’re providing a fast, low-latency, highly visible service across the planet. Our services are as useful to a lone photographer as they are to an agency or a hotel, or an international enterprise. It feels like it should be impossible, yet all the tools and infrastructure are there — we just have to choose extremely carefully. MongoDB Atlas is the only database on our map.”
Get started with MongoDB Atlas today.