Snapchat is a popular social media app that allows users to send disappearing photo and video messages called “snaps”. Since launching in 2011, Snapchat has grown to over 300 million daily active users. One key feature of Snapchat is that by default, snaps disappear after being viewed by the recipient. This ephemeral nature is part of what made Snapchat unique compared to other social platforms.
A common question that arises is whether Snapchat has its own servers to store user data like photos and videos. The short answer is yes, Snapchat does maintain its own internal servers and infrastructure to run its services. However, the full explanation involves understanding how Snapchat stores data as well as its server setup and evolution over time.
How Snapchat Stores Data
When a user sends a snap, the content does not get immediately deleted from Snapchat’s servers after being viewed. There is a backend retention period where snaps are stored temporarily before being deleted. This allows features like replay to work, where recipients can view a snap again within the first 24 hours.
Specifically, snaps are stored encrypted in Snapchat’s database for as long as needed to support replay and other capabilities. Once the retention period expires, the encrypted snap is permanently deleted. Even during this temporary storage period, the content remains inaccessible to Snapchat employees due to encryption. Snapchat databases are spread globally across multiple cloud providers.
Some non-ephemeral user data like profiles, friend lists, and saved media is stored indefinitely until a user deletes their account. Overall, Snapchat does not retain user data very long compared to other social networks. But they still require server capacity to temporarily store snaps and support account features.
Snapchat’s Server Infrastructure
In the early days of Snapchat, the service was supported primarily by Google’s cloud platform. As Snapchat started scaling exponentially in 2013-2014, they began building their own internal infrastructure.
Snapchat now relies on a mix of its own proprietary servers along with leveraging third-party cloud infrastructure. Snapchat has described their system as a globally distributed “cloud-based service”. This means they do not own all their own data centers, but rely on multiple public cloud platforms.
To support their apps and services, Snapchat operates hundreds of thousands of servers across multiple cloud providers. Some estimates indicate Snapchat may have over 200,000 servers across Google Cloud, AWS, and Microsoft Azure. However, their proprietary servers appear to make up over half their infrastructure.
Snapchat has customized server hardware designs optimized for performance, storage, and energy efficiency. For example, they use specialized SSDs and CPUs tailored for Snapchat’s use cases. Extensive software optimization allows Snapchat to efficiently distribute traffic across their mix of proprietary and cloud servers.
Snapchat Server Locations
Snapchat has an extensive global server footprint to provide fast and reliable services worldwide. They have servers in at least 11 countries and territories across North America, Europe, Asia, and Australia.
Some known Snapchat server locations include:
- United States (California, Virginia, Oregon, Iowa)
- Canada (Ontario, Quebec, British Columbia)
- The Netherlands (Amsterdam)
- United Kingdom (London)
- France (Paris)
- Germany (Frankfurt)
- Australia (Sydney)
- Singapore
- Japan (Tokyo)
- Hong Kong
- Brazil (Sao Paulo)
Having servers geographically distributed allows Snapchat to deliver low-latency performance worldwide while also providing redundancy. If servers in one region go down, traffic can be routed to other locations to minimize disruption.
Snapchat’s Custom Server Design
A key part of Snapchat’s infrastructure is the custom hardware powering their proprietary servers. According to their own published specifications, here are some details on Snapchat’s custom server design:
- Uses custom Snapchat motherboard
- Intel Xeon processors (up to 28 cores per server)
- Support for up to 6 terabytes of DDR4 RAM
- Specialized SSD storage up to 64 TB
- 100Gbps networking capacity
- Low-power profile – high core density per rack
Snapchat heavily optimizes the software and workloads running on these servers for efficiency. Their goal is to maximize performance per watt for energy efficiency.
The specialized server hardware lets Snapchat scale up capacity as needed while having more control than completely relying on public cloud infrastructure. Snapchat has the resources to iterate quickly on proprietary hardware suited for their unique use case of ephemeral messaging.
Evolution of Snapchat’s Server Infrastructure
Snapchat’s current server infrastructure reflects years of evolution as the service grew into a global phenomenon:
2011 – Snapchat launches and initially runs entirely on Google Cloud servers. Very minimal infrastructure is needed to support the new startup.
2012 – As the app gains traction, Snapchat shifts some workloads to AWS while remaining primarily on Google Cloud. Server capacity expands gradually.
2013 – Daily active users blow past 5 million as Snapchat’s explosive growth begins. Supporting this requires rapidly scaling up servers.
2014 – By now, Snapchat is pushing Google Cloud infrastructure to its limits. They begin building their own proprietary servers tailored for Snapchat.
2015 – Snapchat opens its first self-owned and operated data center in Iowa. Additional proprietary data centers follow.
2016 – While owning some servers, Snapchat continues leveraging multiple public cloud platforms like Google, AWS, and Microsoft Azure.
2017 – Snapchat is now scaling globally, expanding data centers and servers worldwide to support international growth.
2018+ – Snapchat settles into relying on a mix of its own proprietary servers + public cloud servers. Server hardware and software is heavily customized for Snapchat’s unique use case and performance needs.
This evolution demonstrates how internet services often start small in the cloud, build up proprietary infrastructure as they grow, but still leverage cloud servers for flexibility and redundancy.
Reasons for Snapchat’s Server Infrastructure
There are a few key reasons why Snapchat went from solely using public cloud servers to building hundreds of thousands of custom servers:
- Scale – Snapchat grew bigger than any reasonable public cloud capacity. Their own hardware let them expand servers massively.
- Cost – At their size, owning custom infrastructure became more cost effective long-term versus purely using AWS, GCP, etc.
- Performance – Custom hardware and software optimization unlocks more performance.
- Redundancy – Spread across their own and cloud data centers minimizes disruption if any one site goes down.
- Control – Less reliant on the whims of public cloud providers modifying their services.
Owning and operating a global server fleet does introduce overhead complexities. But for Snapchat, the benefits outweigh the costs given how massive their service has become.
Conclusion
Snapchat clearly has an expansive infrastructure supporting hundreds of millions of users. They’ve evolved from a small startup using Google Cloud to a hybrid model encompassing proprietary data centers, custom hardware, and multiple public cloud platforms.
While the exact breakdown is uncertain, Snapchat appears to rely on roughly 50% proprietary servers and 50% leverage of cloud servers like AWS and GCP. Their global footprint delivers snaps quickly and reliably worldwide thanks to specialized hardware and software. So in summary – yes Snapchat does have its own servers – but it also complements them with public cloud infrastructure for maximum scale, redundancy, and cost efficiency.