Sudden load.
WHAT CAUSES THEM
Marketing activity Being featured somewhere Seasonal events Automated traffic A defective client retrying
WHAT HAPPENS WITHOUT PREPARATION
Response times rise, timeouts cascade, and the service becomes unavailable.
WHAT PROTECTS
Rate limiting Caching aggressively Queueing work rather than performing it immediately Shedding load deliberately
WHAT SHEDDING LOAD MEANS
Rejecting some requests quickly, so the rest succeed.
WHY THAT IS BETTER
Serving most users is better than serving none slowly.
WHAT TO PREPARE IN ADVANCE
Knowing where your limit is Knowing what to disable under pressure Being able to scale, if the architecture allows
WHAT TO DISABLE UNDER PRESSURE
Expensive non-essential features: recommendations, analytics, search suggestions.
WHAT TO CACHE MOST AGGRESSIVELY
Whatever is requested most and changes least.
WHAT TO TEST
Behaviour at several times expected load.
WHAT TO WATCH DURING A SPIKE
Error rate and response times, not only request volume.
WHAT TO AVOID
Discovering your capacity during an event.
