Flipkart Marketplace Platform  ·  View 22 of 25  ·  6 · Operations

Peak Load and Graceful Degradation

The load-shedding contract, agreed in advance so nobody has to invent it at 02:00.

Editable source SVG draw.io All views
Normal day 3x traffic 10x sale hour Brownout Search Full ranking p95 280 ms Scale replicas Cached top queries Lexical only no vector Product page Live price and stock Edge cache 30 s Edge cache 120 s Static shell + stock call Cart and checkout Full flow p95 500 ms Dedicated node pool Virtual waiting room fair queue Protected — never shed Payments All instruments PSP health routing Drop slow PSPs Protected — never shed Recommendations Personalised Shorter candidate set Precomputed segments Off Reviews and Q&A Live counts Cached counts Read only Off Seller analytics Near real time 60 s 5 min refresh Hourly refresh Yesterday only Notifications All classes Priority queue Transactional only Order and payment only Peak Load — What Must Work and What May Break Application we own Data store Decision point Interface / broker Security / platform Opportunity Risk / gap Queue / topic This table is the load-shedding contract, and it is enforced by feature flags rather than by judgement at 02:00. Yellow is a deliberate downgrade; red is switched off. v 1.0 · owner SRE + Product · date 2026-09

The contract

  • Cart, checkout and payment are protected and never shed — every other capability has a written downgrade
  • Each cell is a feature flag with an owner, exercised in the pre-sale game day, not a description of intent
  • Recommendations, reviews and personalised banners go off before anything on the money path slows down

Techniques behind the table

  • CDN and edge caching, pre-warmed caches, precomputed segments and read replicas absorb the read amplification
  • A virtual waiting room makes the queue fair and bounded rather than letting the load balancer decide
  • Queue-based load levelling with backpressure: writes are accepted quickly and processed asynchronously wherever the customer does not need a synchronous answer

Risks

  • Degradations interact: cached prices plus a stale availability filter produces a Buy Box that cannot be bought
  • The waiting room is the most visible failure in the set — if it is wrong, the sale is the story
  • Physical fulfilment capacity, not software, is the binding constraint above about 5x