Question

A national retailer runs its customer-facing web platform on Compute Engine managed instance groups behind a global external HTTP(S) load balancer. Traffic is steady most of the year, but during a two-week holiday sale the site experiences roughly 8x normal load, with sharp, unpredictable spikes tied to flash promotions announced on social media. During last year's sale, autoscaling reacted too slowly and customers saw elevated latency and timeouts for several minutes after each promotion went live. Leadership wants the architecture to meet a 99.95% availability target and absorb sudden spikes without over-provisioning capacity year-round. Which approach best meets these performance and reliability requirements?

Accepted Answer

Combine load-balancing serving-capacity autoscaling with a scale-in control (stabilization) window, and create scheduled scaling to pre-warm a baseline of instances ahead of each announced promotion.

Answer

Configure autoscaling based only on average CPU utilization with the default cool-down period, and increase the maximum instance count for the holiday period.

Answer

Switch the instance groups to Spot VMs during the holiday season to reduce cost and allow a much higher maximum instance count for spikes.

Answer

Replace the managed instance groups with a single very large machine type sized for peak holiday traffic to eliminate autoscaling lag.