Skip to main content
Platform blog

Acceleration Unlocked: DS3_v2 Instance Types on Azure now supported by Photon

New instance types improve price-performance between 2x and 3x for standard benchmarking workloads.
Zach Christopherson
Han Xiao
Desmond Cheong
Kent Marten
Share this post

At Databricks, we offer maximal flexibility for choosing compute for ETL and ML/AI workloads. Staying true to the theme of flexibility, we announce support for additional instance types with Photon on Azure, including default VMs. These choices allow customers to improve performance while reducing the total cost of ownership (TCO). The newly supported VMs include Standard_DS3_v2, which has been added to the growing list of supported VMs. With these instance types, Photon reduces your TCO, making data engineering, analysis, and ML/AI workloads faster. So you pay less overall for performance.

Improving Price-Performance with Photon Acceleration

Since the GA of Photon, the Databricks engineering team has focused on broadening expression and operator coverage in Photon with each release. We have worked on improving the memory utilization of expressions and operators along with the ability to spill when there is memory pressure. With these improvements and optimizations, Photon can now be leveraged against general purpose VMs without losing stability. Photon can be used on a wide spectrum of workloads ranging from raw data processing to optimized execution of queries on highly curated and structured data from BI dashboards. The expansion of supported instance types improves price-performance across all clouds.

Let's look at the data before and after enabling Photon with Databricks Runtime 12.2 using the newly supported instance types.

3.4x and 4.3x relative speed-ups with Photon enabled on DS2_v2 instances.
3.4x and 4.3x relative speed-ups with Photon enabled on DS2_v2 instances.


Relative Price-performance gains of 2.1x and 2.6x with Photon enabled on DS3_v2 instances.
Relative Price-performance gains of 2.1x and 2.6x with Photon enabled on DS3_v2 instances.

Summary of Results

Photon consistently delivers price-performance advantages across TPC-DS and TPC-H derived benchmarks. You see clear price-performance benefits and workload acceleration. TCO gains improved between 2.1 and 2.6x, while the performance boost measured 3.4x to 4.3x using standard benchmarks.

What's Next for Instance Type Support?

We continue to invest in Photon to lower our customers' total costs. Stay tuned as we aim to bring Photon support to a broader range of compute-optimized instances.

Want to learn more about how customers are leveraging the Photon to reduce their costs by more than 30%? Check out this customer-written blog from Vizio.

Try Databricks for free

Related posts

Company blog

How DuPont achieved 11x latency reduction and 4x cost reduction with Photon

This is a collaborative post from Databricks and DuPont. We thank Basant Aggarwal (Lead Enterprise Architect), Romain Fardel (Senior Data Scientist), and John...
Platform blog

Announcing the General Availability of Predictive I/O for Reads

Today, we are excited to announce the general availability of Predictive I/O for Databricks SQL (DB SQL) : a machine learning powered feature...
Platform blog

Announcing the Public Preview of Predictive I/O for Updates

Previously, we’ve shown you how a new technology called Predictive I/O could improve selective reads by up to 35x for CDW customers without...
See all Platform Blog posts