Run a personalised film campaign.
Keep compute costs under control.
For T&DA’s Cadbury campaign, Quokka helped combine AWS native services with on-premises GPU compute. Cloud burst capacity handled overflow as demand grew, helping control cost while meeting the processing load.
T&DAThe challenge
A personalised film campaign needs enough compute to create each film while people are taking part. Rendering time, concurrent jobs, storage and repeat viewing all affect the capacity and budget behind the experience.
What we brought
Quokka helped T&DA combine native AWS services with Elastic Container Service workloads on premises and local GPU compute. The architecture used cloud capacity for burst overflow when load called for more performance. We modelled processing capacity alongside storage and delivery costs.
Additional cloud processing when demand exceeds local capacity.
Plan for the workload, then the cost.
Processing capacity
The compute model connects film volumes, render time and concurrent GPU jobs to the available processing window. It compares full-day capacity with the shorter period when people submit requests, so the team can assess the effect on queues.
Storage and delivery
The media model accounts for new films, accumulated storage, repeat playback and downloads. It gives T&DA a way to examine how participation and viewing behaviour affect the campaign’s operating costs.
Scale out when load requires it
The campaign used on-premises GPU capacity for the regular workload and cloud compute for burst overflow. Scaling out when processing demand increased helped control costs while adding performance when needed.
The creative work
T&DA developed the personalised video experience for Cadbury’s “A Cheer For Every Volunteer” campaign. Its campaign credits name Mondelez as client and Ogilvy and Hogarth as agency.
Explore T&DA’s campaignWhat needs to work better?
Talk to us about the workload, the operating constraints and the cost.