Kyvos on Microsoft Azure
Applies to: Kyvos Enterprise Kyvos Cloud (SaaS on AWS) Kyvos AWS Marketplace
Kyvos Azure Marketplace Kyvos GCP Marketplace Kyvos Single Node Installation (Kyvos SNI)
Kyvos on Microsoft Azure with No-Spark
The diagram illustrates Kyvos’ No-Spark architecture, deployed on Microsoft Azure, which is purpose-built for high-performance, scalable analytics without dependence on Spark-based frameworks. The architecture integrates a wide array of data sources, including Apache Iceberg, Parquet, CSV files, Azure Data Lake Storage, Azure Synapse, and Snowflake.
At its core, the Elastic Processing Cluster and Elastic Querying Cluster—both deployed using Azure Virtual Machine Scale Sets (VMSS)—enable dynamic scalability and distributed computing for semantic model processing and query execution. These components interact with the Kyvos Data Storage layer, which holds both raw and aggregated data. The Analytical Server, hosted on an Azure VM, manages orchestration tasks like job scheduling, security, and metadata handling.
Positioned above these layers is the Semantic Layer, which simplifies data interactions for business users by providing a unified and performance-optimized semantic view. Kyvos supports a wide range of query languages such as SQL, MDX, DAX, and LangChain, along with REST and Java APIs for programmatic access.
End users can consume insights via leading BI tools like Tableau, Power BI, Looker, Excel, and Strategy, in addition to Kyvos-native interfaces including Kyvos Dialogs, Kyvos Viz (powered by AI), the Kyvos Excel Add-in, and Kyvos Reporting. To ensure sub-second query performance, the Smart Aggregate Cache stores frequently accessed aggregates.
This Azure-native architecture is designed for elasticity, high concurrency, and enterprise-grade performance while also integrating advanced AI-powered capabilities for enhanced analytics experiences.
Kyvos on Microsoft Azure with Spark
Kyvos brings the power of multidimensional analytics to Azure. You can process a modern BI architecture on Azure with built-in elasticity and perform complex, multidimensional analytics on your cloud workloads with unmatched performance and unlimited scalability.
Kyvos consists of two main components: BI Servers and Query Engines. Kyvos BI servers are deployed on standalone Azure Virtual Machines (VMs), and the query engines are deployed on VMs in Virtual Machine Scale Sets. Query engines capacity can be configured to increase or decrease depending on the load.
Auto-scaling enables Kyvos to scale up and down query engines at the time of semantic model processing using Databricks on Azure.
Kyvos reads data from Azure Data Lake Storage (ADLS) and processes it using Databricks. It launches a series of Spark jobs for semantic model processing.
At the time of Kyvos deployment, leveraging the Databricks service, you can either provide a fixed number of worker nodes for the cluster or define the minimum and a maximum number of worker nodes. The cluster then scales in or out to use only the needed resources.
Databricks chooses the appropriate number of worker nodes required to run your job. This ensures that only the required number of machines are used during the semantic model processes.
Once the semantic models are processed, they are stored in ADLS GEN 2 for persistent storage. This helps deliver much higher performance as compared to querying directly from Blob storage.
Kyvos supports querying elasticity through scheduled scaling. Based on the expected loads, you can specify the day/time when resources need to scale up or down. This helps reduce costs during lean periods.
Kyvos uses ADLS GEN2 shared storage to store and cache semantic model data.
Kyvos' modern architecture enables deep integration in the Azure ecosystem. As shown in the following figure, Kyvos also supports the Snowflake data warehouse on Azure.
Related topics