Everything I learned about dynamic scaling of cloud apps.
summary
A live One Direction event drives massive ten‑minute spikes so the team uses App Engine for VM autoscaling but avoids constant capacity costs by writing numeric metrics into dedicated memcache with sharded counters, tracking only top‑30 countries plus a rest‑of‑world bucket, and keeping a 32-entry hero list for winners. The design sacrifices full raw records for fast increments, preserves privacy, handles tall skinny spikes, and yields a cloud bill smaller than the post‑event drinks.
AI-written summary · about
chapters (7)
- 0:00Start
- 4:01Compute Capacity
- 4:47Promise of the Cloud
- 7:34No Sequel
- 8:14Memory Caching
- 11:09Sharding
- 17:22Scalability
related talks
44mIntroduction to application scaling
53mWhat I Learned About Going Fast at eBay & Google
49mWhat I Learned About Going Fast at Ebay & Google
18mScaling APIs to Billions: Extreme-Scale Reliability
01h 03mThe Trail to Scale Without Fail: Rails?
47mBuilding Scalable Web Apps
47mScaling: from 0 to 25 million users
30mAbove the Clouds: Investigating Systems from 32,000 ft.