See Rockset
in action

Get a product tour with a Rockset engineer

Rockset vs Apache Druid

Compare and contrast Rockset and Apache Druid by architecture, ingestion, queries, performance, and scalability.

Rockset vs Apache Druid Architecture

Apache Druid
Deployment model
SaaS - infrastructure, software and cluster ops managed by service provider
• SaaS or self managed. • Imply’s Druid cloud offering requires customers to configure, scale, and capacity plan
Use of storage hierarchy
• SSDs store shared hot data, accessible from any Virtual Instance cluster • Cloud storage for durability
• Queries are served from data on disk and an in-memory cache • Cloud storage or HDFS for deep storage
Isolation of ingest and query
Yes - separate compute clusters (Virtual Instances) for ingest and query
• Ingestion and queries are run on the same node by default • The ingestion and querying processes can be run on separate nodes, however not for real-time data
Separation of compute and storage
No, although Imply’s offering separates compute and storage
Isolation for multiple applications
Yes - separate compute cluster (Virtual Instance) for each application

Rockset is built to be a cloud-only database and does not have a self-managed option. It disaggregates compute from both hot storage and cloud storage, allowing multiple isolated compute clusters to run on the same shared data.

Druid’s architecture employs nodes called data servers that are used for both ingestion and queries. High ingestion or query load can cause CPU and memory contention compared with Druid alternatives. Breaking apart the pre-packaged ingestion and query server components involves planning ahead and additional complexity, and is not dynamic.

Rockset vs Apache Druid Ingestion

Apache Druid
Data sources
Managed data connectors to: • Events streams (e.g. Kafka, Kinesis) • Database CDC (e.g. MongoDB, DynamoDB, MySQL, PostgreSQL) • Data lakes (e.g. S3, Google Cloud Storage)
Data connectors to: • Events streams (e.g. Kafka, Kinesis) • Data lakes (e.g. S3, Google Cloud Storage) • RDBMS and HDFS databases CDC events from databases require manual conversion to Druid events
Semi structured data
• Ingests JSON and XML without a predefined schema • Ingests nested data
Druid requires flattening nested data at ingest and maintaining a flattening spec as the schema changes over time
Transformations and rollups
Yes - using SQL ingest transformations
Yes, using ingestion specs written in JSON support rollups and simple transformations. SQL ingest transformations available for Imply

Rockset has built-in connectors that manage streaming ingestion from common data sources. It has native support for semi-structured data, so that nested JSON and XML can be ingested and queried as is.

Druid has built-in connectors that manage ingestion from common data sources. Unlike some Druid competitors, it doesn’t support nested data, so data must be flattened at ingest. Denormalization is also required at ingest, increasing operational burden for certain use cases.

Rockset vs Apache Druid Performance

Apache Druid
Documents are mutable and can be updated at the field level
Druid segments become immutable once committed and published, making it more appropriate for append-only use cases
• Converged Index (row, columnar and inverted index) built on all data by default • All queries are resolved through the index
• Bitmap index
Query latency
50-1000ms queries on 100s of TB
Typically sub-second query latency for denormalized, flattened datasets up to 100s of TBs
Storage format
Converged Index, comprising a rowstore, columnstore and inverted index
Columnar format partitioned by time
Streaming ingest
• Ingests on a per-record basis • Data latency is typically 1-2 seconds
• Sub 15 seconds

Rockset is designed to make streaming data queryable as quickly as possible by avoiding the need to batch data. It also updates documents efficiently by only reindexing fields that are part of an update request. Rockset indexes all data by default, which results in storage amplification but also enables low-latency queries that require less compute.

Druid is designed to make streaming data queryable as quickly as possible. JOINs are either impossible or incur a large performance penalty. Updates are only possible via batch jobs. Druid leverages data denormalization and write-time aggregation at ingestion to reduce query latency.

Rockset vs Apache Druid Queries

Apache Druid
Yes, only for datasets that fit in memory, and with a query latency penalty
Query language
SQL - with extensions for nested data
Druid native queries • Druid SQL
Developer tooling
• Data APIs - saved SQL queries executed via REST endpoint • Python, Java, Node.js and Go SDKs • UDFs
• Druid SQL API
Visualization tools
Integrations with Tableau, Looker, Grafana, Superset, Power BI, Retoolwhic
Pivot, maintained by Imply

Rockset supports SQL as its native query language and can perform SQL joins. Users can create data APIs by storing SQL queries in Rockset that are executed from dedicated REST endpoints. Rockset integrates with some common visualization tools, but BI is not Rockset’s primary use case.

Druid has a native JSON-based query language and provides Druid SQL as an alternative that translates into its native queries. JOINs are not recommended.

Rockset vs Apache Druid Scalability

Apache Druid
Vertical scaling
Resize compute clusters (Virtual Instances) via API or in the console
Users can manually increase or decrease server sizes
Horizontal scaling
• Add or remove compute clusters (Virtual Instances) via API or in the console • Scale out compute clusters for higher concurrency • Use separate compute clusters to isolate ingest from query or for multiple isolated applications • No rebalancing required
• Users can manually add additional nodes to a scaled-out cluster. • Imply automates some of the horizontal scaling process.

Rockset Virtual Instances are distributed compute clusters that can be scaled up for faster queries or scaled out for practically unlimited concurrency or if compute isolation is needed. Rockset has shared storage that scales automatically and independently, so no rebalancing is required.

Druid users are exposed to complex decisions about the number and size of servers as clusters are scaled.