What this blog covers:
- Find how Kyvos compares with Starburst for query cost, performance and other parameters.
- Read what difference Kyvos’ AI-powered semantic layer and advanced OLAP features can make to a modern data stack, compared to Starburst’s limited functionalities.
- Understand how Kyvos outranks Starburst on all counts.
Starburst provides a data lakehouse solution powered by Trino, a SQL analytics engine. It is a compute-intensive, in-memory solution. The platform has limited functionalities, offers slower query responses and incurs a much higher cost of ownership to serve enterprise-wide BI and analytics needs.
Kyvos is a real universal semantic layer, purpose-built to simplify analytics and speed up query performance for large scale enterprise-wide data processing, without any constraints of speed, scale or cost.
This blog compares Kyvos and Starburst in terms of features and performance.
Responses for Complex Queries
Starburst is an in-memory solution, which leads to scalability and performance issues when enterprise data volumes and complexities grow. As per benchmark reports, Starburst logged a query response of 15 seconds for a single user on 1 TB data using 8+1 nodes of 24 cores and 256 GB. Comparatively, Kyvos has a query response of 1 second for a single user on 1.5 TB data using 11 nodes of 8 cores and 64 GB.
Kyvos employs intelligent aggregation and caching, enhanced by AI-driven self-tuning to efficiently handle large data volumes and query demands. Combined with load-balancing, the platform enables sub-second query responses on billions of rows, without any loss in performance for 1000+ concurrent users.
Querying Cost
Being an in-memory solution, high-configuration nodes are required on Starburst. This results in a manifold increase in the cost of deployment, estimated to be 4-5 times or more than Kyvos on 5 billion records with 1-1.5 TB data size.
For data refresh with a new dataset, there is no additional infra deployment needs with Kyvos, while new datasets require additional spending on compute and memory with Starburst.
Query Engine
Starburst uses Trino’s in-memory, distributed query engine. It is impaired with performance issues when handling higher data volumes having multiple tables and large number of columns. On the other hand, Kyvos’ optimized query engines run OLAP queries on multidimensional data models with no performance issues, even with large datasets.
Semantic Layer Capabilities vs. Data Lakes Performance
Starburst data lake products only partially work as a semantic layer. They do not meet the requirements of a unified and standardized view with large data sets, multiple departments and a large number of users. It requires creating multiple SQL queries and views to support data at scale, leading to high maintenance costs.
Kyvos supports thousands of fields, billions of rows and dozens of levels of drilldowns in a single semantic model. It creates centralized calculations only once for all data sources, and they can be used across multiple BI/ analytics tools. However, Starburst creates multiple SQL queries and views to support data at scale, leading to high maintenance. Also, users have to write SQL queries for every data source, except Hive and Iceberg, on Starburst.
Kyvos supports advanced multidimensional data modeling with dimensions, hierarchies and measures easily defined through an intuitive UI. Starburst instead has a relational data modeling approach that works only with SQL.
Aggregation Tables and Caching
Kyvos creates AI-powered smart aggregates based on query patterns. Data models are created once and used multiple times. Starburst supports materialized views only for Hive and Iceberg data sources. Due to these limitations, organizations need to create several aggregate tables that are uncontrolled and very difficult to manage.
Apart from this, Starburst supports caching and indexing through Starburst Warp speed, however only for Hive, Iceberg and Delta Lake. Kyvos supports multi-level cache for all data sources.
Advanced OLAP Features
Since SQL is the only query language on Starburst, advanced OLAP functions, such as Scope, Alter Cube, Default Member, Cubevalue, etc., that are available with DAX and MDX functions are not supported. In addition, it does not have out-of-the-box support for unbalanced and ragged hierarchies or parent-child and alternate hierarchies with custom rollups.
All these advanced financial hierarchies and data modeling capabilities are an integral part of the Kyvos offerings.
Gen AI Capabilities
Kyvos leverages Gen AI with natural language summarization that goes a long way in making insights and analytics accessible for business leaders. Moreover, it empowers them with guided conversational data exploration with support for prompts, persistent contexts, backtracking and parallel explorations. Starburst offers none or has very limited Gen AI capabilities with support only for SQL, but not extended to MDX query generation.
Conclusion
Compared to Starburst, Kyvos stands out as a comprehensive and efficient solution for comprehensive and scalable data analytics. Kyvos’ AI-powered semantic layer supports advanced multidimensional data modeling and centralized calculations, simplifying data analysis across various BI tools. Furthermore, advanced OLAP features and Gen AI capabilities make it a versatile, more cost-effective and performant solution for enterprises looking to handle vast data volumes and complex queries.
Access the full report and a detailed cost and performance comparison between data architectures for analytics using Kyvos and Starburst here.