Apache Solr deep dive: From full-text search to distributed search architecture

In the first part of our Search Trilogy, we explored Azure AI Search and how Microsoft combines traditional search, AI enrichment, and vector-based retrieval into a managed cloud solution.

In this second part, we shift the focus to Apache Solr — one of the most widely adopted enterprise search platforms built on Apache Lucene. While Azure AI Search provides a managed cloud-native experience, Solr offers deep flexibility, advanced search capabilities, and full control over distributed search infrastructure.

Apache Solr powers search experiences for organizations such as Netflix, Disney, Bloomberg, eBay, SAP Hybris, and Instagram, proving itself a mature, highly scalable search engine for enterprise environments.

Why Apache Solr?

Traditional SQL databases are not optimized for modern search workloads.

Although SQL queries can support simple filtering and text matching, enterprise applications often require:

• Full-text search
• Multi-language support
• Synonyms and stemming
• Relevance ranking
• Typo tolerance
• Distributed search
• Extremely fast retrieval across large datasets

This is where Apache Solr becomes valuable.

Unlike relational databases, Solr is designed specifically for document retrieval and large-scale search operations.

Key advantages include:

• Advanced full-text search
• Relevance scoring with BM25 and TF-IDF
• Faceted search
• High availability
• Horizontal scalability with SolrCloud
• Efficient inverted indexes
• Fast search performance even with billions of documents

Solr is especially suitable for:

• E-commerce platforms
• News and media portals
• Documentation systems
• Knowledge bases
• Product catalogs
• Large enterprise search systems

Apache Solr architecture

Apache Solr is built on top of Apache Lucene.

Lucene provides the low-level indexing and retrieval engine, while Solr adds enterprise-grade features such as:

• HTTP APIs
• Distributed search
• Clustering
• Schema management
• Analytics
• Monitoring
• High availability
• Faceting and filtering
• AI-related search features

At a high level, the architecture consists of:

• Solr nodes
• Collections
• Shards
• Replicas
• ZooKeeper
• Lucene indexes

Core components

How Apache Solr works

Indexing flow

The indexing process begins when a client sends documents to Solr using the /update endpoint.

Example flow:

• Client sends document
• Solr transforms it into a SolrInputDocument
• Text analyzers process the content
• Lucene creates inverted indexes
• Solr commits changes and makes data searchable

During analysis, Solr applies:

• Tokenization
• Lowercasing
• Stopword removal
• Synonym handling
• Stemming

Example:

"Electronics and electronic devices"
→ ["electronic", "device"]

This processing significantly improves search relevancy.

Searching flow

Search requests are typically sent through HTTP APIs.

Example query:

/solr/products/select?q=electronics&fq=category:phones&sort=price asc

The process:

1. Request Handler processes the query
2. Solr receives a request
3. Query Parser converts input into Lucene query objects
4. Filters are applied separately for caching efficiency
5. Lucene performs a search against indexes
6. Results are ranked and returned

Solr supports multiple query parsers:

The fq parameter is especially important because filter queries are cached independently, improving performance for repeated searches.

Solr vs SolrCloud

Standalone Solr works well for development and smaller systems.

However, enterprise production environments usually require:

• High availability
• Fault tolerance
• Distributed indexing
• Distributed querying
• Horizontal scaling

This is where SolrCloud comes in.

SolrCloud transforms multiple Solr nodes into a distributed search cluster.

Key concepts:

• Data split into shards
• Each shard has replicas
• ZooKeeper manages cluster coordination
• Automatic failover support
• Distributed querying across nodes

This architecture enables massive scalability and resilience.

AI features in Apache Solr

Modern search platforms increasingly integrate AI and semantic search capabilities.

Apache Solr supports AI-oriented search scenarios through:

Although Solr does not provide built-in AI enrichment like Azure AI Search, it offers strong flexibility for custom AI pipelines.

Practical example: Running SolrCloud with Docker Compose

One of the easiest ways to start experimenting with distributed search is through Docker.

The following example creates:

• One ZooKeeper node
• Two Solr nodes
• Shared Docker network
• Persistent storage volumes

version: "3.8"

networks:
  solr-net:
    driver: bridge

volumes:
  zk-data:
  solr1-data:
  solr2-data:

services:
  zk:
    image: zookeeper:3.8
    container_name: zk
    hostname: zk
    environment:
      ZOO_MY_ID: 1
      ZOO_CLIENT_PORT: 2181
      ZOO_TICK_TIME: 2000
      ZOO_4LW_COMMANDS_WHITELIST: mntr,conf,ruok
    volumes:
      - zk-data:/data
    networks:
      - solr-net
    ports:
      - "2181:2181"

  solr1:
    image: solr:9.0
    container_name: solr1
    hostname: solr1
    command: solr -f -c -z zk:2181
    ports:
      - "8983:8983"
    volumes:
      - solr1-data:/var/solr
    environment:
      - SOLR_HEAP=1g
      - SOLR_HOST=solr1
    networks:
      - solr-net

  solr2:
    image: solr:9.0
    container_name: solr2
    hostname: solr2
    command: solr -f -c -z zk:2181
    ports:
      - "7574:8983"
    volumes:
      - solr2-data:/var/solr
    environment:
      - SOLR_HEAP=1g
      - SOLR_HOST=solr2
    networks:
      - solr-net

To start the cluster:

docker compose up -d

Once started:

• Solr node 1 → http://localhost:8983
• Solr node 2 → http://localhost:7574

This setup provides a practical introduction to:

• SolrCloud
• Distributed indexing
• Replication
• ZooKeeper coordination
• Multi-node search clusters

Final thoughts

Apache Solr remains one of the most powerful and mature enterprise search platforms available today.

Compared to traditional SQL search, Solr delivers:

• Better relevancy
• Faster retrieval
• Distributed scalability
• Advanced text analysis
• AI-ready search capabilities

Compared to Azure AI Search, Solr offers:

• Greater infrastructure control
• Open-source flexibility
• Custom deployment options
• Deep query customization

At the same time, it also requires more operational ownership, especially around cluster management, scaling, and maintenance.

For organizations building:

• Large-scale search systems
• E-commerce platforms
• Knowledge retrieval systems
• AI-enhanced search applications

Apache Solr remains a highly relevant and battle-tested solution.

In the next part of the Search Trilogy, we will explore Elasticsearch and compare its architecture, indexing model, distributed capabilities, and AI ecosystem with both Azure AI Search and Apache Solr.