What is data virtualization?

Data virtualization is a solution that creates intermediate layers between data consumers and disparate data source systems. These systems give consumers a unified interface to all data in the organization while leaving the data itself at the source. 

This approach eliminates the need to move or copy data through ETL into centralized data repositories.

What are the benefits of data virtualization?

As cloud technologies have enabled the era of big data, the problem of data sprawl has gotten more severe. Attempts to centralize analysis through data warehouses and data lakes struggle to scale as ETL maintenance requirements run into resource limits.

Creating a virtual data layer that leaves data at the source while supporting agile analysis has been a long-held data management dream. Only recently, however, have companies started to see data virtualization’s benefits.

Below we highlight 10 data virtualization benefits. 

1. Cloud computing enabled data virtualization

The move to the cloud has both enabled data virtualization and allowed virtualization to make cloud-based big data better.

Old monolithic architectures tightly coupled storage and compute while prioritizing on-premises network efficiency. The cloud disaggregates storage from compute while making network costs negligible.

The cloud’s scalable, elastic infrastructure and cost-efficient economic model are perfectly tailored to data virtualization. Data stays in place at its source in cloud, multi-cloud, and hybrid-cloud architectures. Virtualization platforms spin up compute nodes to run queries on-demand for customers who do not need to know how or where the data is stored.

2. Data virtualization does not replace existing centralized repositories

Another advantage of data virtualization is that it does not replace existing centralized repositories. Companies developed each of these data stores for reasons that are still valid. The only reason existing data warehouses or data lakes become cumbersome is that companies make them do too much in the vain pursuit of a single source of truth.

By tying these sources together within unified, virtual views of the company’s data, virtualization systems decouple data analysis from storage infrastructure. Data teams can design and optimize their repositories without impacting the company’s data users.

Freed from understanding where and how different data are stored, companies get a single point of access to support their data-driven cultures.

3. Creates business agility and more effective data-driven decision-making

Data virtualization speeds time-to-insight and supports faster, more effective data-driven decision-making. Business intelligence analysts can query data sources directly rather than waiting for data engineers to develop and test new ETL pipelines. 

In turn, data engineers spend less time on ETL maintenance which frees them to support data scientists working on machine learning and other complex projects.

4. Increased scalability and elasticity

By supporting on-demand, self-service analysis, data virtualization makes a company’s data architecture more scalable and elastic. Companies don’t need to over-invest in computing capacity. 

Instead, data virtualization lets companies leverage DevOps practices to scale cloud compute up and down programmatically as needed.

5. Data integration and accessibility

Data virtualization platforms can integrate many disparate data systems, flat files, databases, and API-based sources, regardless of their underlying technology. At the same time, virtualization erases the data silos and domain-specific data structures or formats that keep datasets opaque.

Simplifying integration improves data accessibility by making technological implementations transparent to data users.

6. Single point of access

Data virtualization systems remove the complexities of querying multiple data sources. Behind the scenes, query engines handle the type of data and format each source uses.

Data customers see a consistent, unified data layer where they can access data through a user-friendly interface.

7. Real-time global data access

By implementing data virtualization, companies let users source data no matter where it physically resides. Direct access means users no longer wait for data teams to design or update ETL pipelines.

Business intelligence analysts can produce results faster with the most current information.

8. Data abstraction 

As data volumes and velocities continue to grow, valuable data gets stored across a mix of cloud storage engines and relational and non-relational databases. Querying each source directly requires knowledge of the data and a deep understanding of each storage type’s structure.

Data virtualization provides an abstraction layer that hides that complexity to simplify access to heterogeneous data sources.

9. Data virtualization delivers cost savings

Data virtualization delivers cost savings by turning many capital expenditures into operational expenses. Leaving data at its source rather than consolidating data in warehouses reduces storage costs. Disaggregating storage from compute lets companies optimize their storage costs while shifting their computing expenses to a pay-as-you-go model.

At the same time, eliminating data engineers’ burden of ETL maintenance reduces hiring pressures as the business scales.

10. Enhanced analytics capabilities

Creating an abstraction layer to integrate structured and unstructured data sources streamlines advanced data analytics such as data mining and machine learning initiatives.

What is the difference between ETL and data virtualization?

ETL pipelines and data virtualization are different approaches to integrating and managing data to feed data services and analysis. 

The ETL process extracts data from the source, transforms the data to meet the product’s formatting and other requirements, and loads the data into the next storage location.

Data virtualization creates an abstraction layer that collects metadata from each source to present to users. The system applies any needed transformations in real time as queries run.

These approaches have different implications for how companies manage data.

Data movement

ETL: These pipelines move or copy data from a source towards a data warehouse, increasing duplication in the process.

Data virtualization: All data storage remains at the source. Rather than moving or copying data to another location, data virtualization software feeds data to temporary compute resources that disappear when the query is complete.

Data integration

ETL: Pipelines typically extract data in batches for transformation before loading it into a data warehouse.

Data virtualization: Data integration from multiple sources happens in near real-time for presentation to users in a unified view. 

Data storage

ETL: Because ETL pipelines run intermittently, they typically copy or move the data to a data warehouse for further analysis.

Data virtualization: Data duplication is unnecessary since it remains at the source. Services can use the virtualization layer for real-time data access.

Data latency

ETL: Users must wait for ETL pipelines to refresh a data warehouse before beginning their analysis. This time lag could mean they are not working with the most current data possible.

Data virtualization: Since users can query data sources directly without waiting for pipeline updates and schedules, they can always access the most up-to-date information.

Data governance and security

ETL: Each pipeline must implement data governance and security policies, increasing development and maintenance burdens.

Data virtualization: Within the virtual data layer, companies can programmatically implement and monitor governance policies that control access and data quality.

What is the difference between data visualization and data virtualization?

Mixing up data visualization and data virtualization is easy to do when speed typing or dealing with an overly-helpful auto-correct system.

Data visualization is the graphical presentation of data in an analysis or real-time monitoring system.

Data virtualization is the creation of a virtual data layer that abstracts the technical details of disparate storage systems to give data consumers easy, direct access to data at the source.

Data virtualization made better with Starburst 

Starburst delivers a single point of access to every data source, unlocking valuable data where it lives and enabling fast, accurate analysis and business-driving insights.

With connectors for more than 50 enterprise data sources, Starburst quickly integrates data sources across your storage infrastructure. Over time, Starburst’s multi-platform support gives your company the optionality you need to avoid vendor lock-in.

Starburst leverages the Trino open source-based massively parallel processing (MPP) engine to performance-enhancing features that reduce query times and speed time-to-insight. One customer experienced a 10-20X improvement in complex SQL queries after integrating its mix of cloud and on-premises platforms through Starburst’s Enterprise Platform.

ETL pipelines and data warehouses are made better with Starburst’s virtual data layer frees resources and duplicative storage expenses. Another customer achieved a 61% reduction in TCO by replacing its centralized storage and ETL pipelines with Starburst’s virtualization approach.

Data is only going to increase in scale, velocity, and value. Chasing after the mythical single source of truth will always be an expensive exercise in futility. Modern data virtualization solutions like Starburst let you leave your data in place while making data more accessible and productive.

Related reading: Starburst Data Virtualization: Delivering data virtualization and federation at enterprise scale