How Send.Gallery Scales to 10,000+ Guests Without Breaking a Sweat
From intimate weddings to massive festivals, our infrastructure handles galleries of any size. Here is the technology and architecture behind the scale.
The scale challenge in event photo sharing
Event photo sharing has an unusual scale profile. A 30-person birthday party might generate 300 photos. A 200-person wedding generates 2,000. A 1,000-person corporate conference generates 5,000. A 10,000-person marathon generates 40,000. A 50,000-person music festival generates 100,000+.
The challenge is not just storing these photos — it is matching faces across them. When a guest takes a selfie, the system needs to compare their face against every face in the gallery. For 100,000 photos, that is potentially millions of face comparisons, and the result needs to come back in under five seconds.
Most photo sharing platforms handle small events well but fall apart at scale. Uploads slow down, matching takes minutes instead of seconds, and the gallery becomes unusable during peak traffic when hundreds of guests are searching simultaneously.
Server-side face matching for speed
Send.Gallery performs all face matching on the server side, not on the client device. This means the comparison happens on powerful servers with GPU acceleration, not on a guest's phone. The phone only sends a single selfie image; the server does all the heavy lifting.
The face matching algorithm uses approximate nearest neighbour search, a technique that finds matching faces in logarithmic time rather than linear time. This means a gallery with 100,000 faces can be searched nearly as fast as one with 1,000 faces. The matching time grows very slowly with gallery size.
In practice, a selfie search across a 40,000-photo marathon gallery returns results in under three seconds. A search across a 100,000-photo festival gallery returns results in under five seconds. The experience feels instant regardless of event size.
CDN delivery for global access
Photos are delivered through a global Content Delivery Network (CDN). When a guest in Mumbai requests a photo, it is served from an edge server in Mumbai. When a guest in London requests the same photo, it is served from an edge server in London. The photos are cached close to the user.
This means download speeds are fast regardless of the guest's location. A full-resolution 8 MB photo downloads in under two seconds on a 4G connection. At events with international attendees — destination weddings, international conferences, global festivals — this global delivery is essential.
The CDN also handles traffic spikes gracefully. When 500 guests at a wedding all scan the QR code within a 10-minute window during the reception, the CDN absorbs the traffic without any degradation in performance. The gallery remains responsive even under heavy concurrent load.
Handling concurrent selfie searches
At a large event, many guests search for their photos at the same time. At a marathon finish line, hundreds of runners might search within a 30-minute window. At a festival, thousands of attendees might search during a headliner's set.
Send.Gallery's matching infrastructure is designed for concurrency. The face matching service runs on multiple servers that can be auto-scaled based on demand. When traffic spikes, additional matching servers come online automatically. When traffic subsides, they scale down.
This elastic architecture means the platform handles 10 concurrent searches or 1,000 concurrent searches with the same response time. The guest experience is consistent regardless of how many other people are searching at the same time.
Real-world scale: festivals and marathons
Send.Gallery has powered over 5,000 events ranging from 30-person parties to 10,000+ participant marathons. The platform has scanned over 200 million faces and delivered over 70 million photos, all with 99.9% matching accuracy.
At music festivals, the scale challenge is compounded by the sheer volume of photos and the desire for real-time delivery. Send.Gallery supports direct camera uploads during the event, so festival-goers can find photos taken earlier in the day while the festival is still going. AI highlight reels compile each guest's photos into a personal video for Instagram.
For marathons, the platform handles the finish-line rush — hundreds of runners searching for their photos within minutes of crossing the finish line. The system returns results in under three seconds per runner, with hundreds of concurrent searches handled without any slowdown. Photo selling at sports events adds another layer of scale, as runners browse, purchase and download photos simultaneously.
Built for the next million events
The architecture is designed to scale well beyond current volumes. As Send.Gallery grows from 5,000 events to 50,000 to 500,000, the infrastructure scales horizontally — more servers, more CDN edges, more storage — without any change to the user experience.
For photographers and event organisers, this means you never have to worry about whether the platform can handle your event. A 50-person birthday and a 50,000-person festival run on the same infrastructure, with the same speed, the same accuracy and the same reliability. Scale is not a feature — it is a foundation.
Back to blog