How DoorDash Optimized Item Availability at Scale Using Elasticsearch

2026-04-10 · Source: HackerNoon · Field: Technology & Digital — Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

DoorDash significantly improved item availability checks for its high-traffic homepage by moving from live menu service calls to an Elasticsearch-based indexing system. Initial attempts to query the menu service directly resulted in high latency (P99 > 300ms) due to fan-out issues and frequent cache misses from dynamic availability data. The team then indexed real-time menu updates into Elasticsearch. They explored three indexing strategies: nested documents, which yielded 600ms P99 latency due to expensive joins; encoded terms, which reduced latency to 350ms but increased index size by 6x; and finally, integer range fields backed by BKD trees, achieving 250ms P99 latency and baseline storage. This optimization was critical for meeting the homepage's sub-300ms rendering budget.

Key takeaway

For Software Engineers building high-traffic, real-time data services, directly querying a transactional source for availability at scale is often unsustainable. You should prioritize indexing dynamic data into a purpose-built search engine like Elasticsearch. Specifically, explore range field types and BKD trees for time-based availability to achieve optimal latency and storage efficiency, validating changes with production traffic replay.

Key insights

Optimizing high-scale data availability requires indexing and careful selection of database field types.

Principles

• Match data access patterns to service capabilities.
• Index frequently changing data for high-volume reads.
• Field type choice significantly impacts query performance.

Method

Index real-time availability updates into Elasticsearch, using integer range fields with BKD trees to represent time windows for efficient range queries.

In practice

• Replay production traffic for search infrastructure changes.
• Consider storage constraints as a driver for optimization.
• Evaluate Elasticsearch field types for specific query needs.

Topics

• Elasticsearch Optimization
• Item Availability
• Real-time Indexing
• BKD Trees
• Latency Reduction

Best for: Software Engineer, Data Engineer, IT Professional

Related on AIssential

• Database Performance Strategies and Their Hidden Costs — Database optimization involves inherent tradeoffs where improving one metric often negatively impacts another.
• Building Real-Time Product Search on Databricks — Modern product search requires an integrated platform for real-time ingestion, retrieval, and refinement, balancing speed with relevance and business outcomes.
• Nextdoor’s Database Evolution: A Scaling Ladder — Nextdoor's database architecture evolved from a single PostgreSQL instance to a sophisticated distributed system to support millions of users, addressing scaling challenges through a disciplined progr…
• Article: Stragglers, Not Failures: How Adaptive Hedged Requests Reduce p99 Latency by 74 Percent — Adaptive hedging, using real-time latency distribution learning and a budget, effectively mitigates straggler-induced tail latency without manual tuning.
• Building a Real-Time Weather Data Pipeline using Kafka, Spark, and Grafana — Modern data pipelines integrate Kafka, Spark, TimescaleDB, and Grafana for real-time data ingestion, processing, storage, and visualization.
• Netflix Serves 84% of Query Results from Cache with Interval-Aware Caching in Apache Druid — Decomposing query results into time-aligned segments enables high cache reuse for rolling window analytics.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by HackerNoon.