Snowflake has announced a new compute service for automatically managing infrastructure. Adaptive Compute automatically selects the right cluster size and number of clusters for specific tasks. The auto-suspend and resume duration is also determined automatically based on the workload.
“Users no longer need to think about the warehouse size, concurrency settings, multi-cluster warehouse settings, or auto-suspend/resume semantics,” Snowflake said in its announcement at the annual Snowflake Summit. Adaptive Compute uses intelligent technology to route queries to the right clusters. Users do not need to take any action.
Snowflake will support warehouses created with the Adaptive Compute service. The company also refers to these as Adaptive Warehouses. All Adaptive Warehouses within an account use a shared resource pool. This should maximize efficiency by utilizing available computing power optimally. Users retain their familiar billing model and granular information via ACCOUNT_USAGE views.
Migration without downtime
The transition from a standard virtual warehouse to an Adaptive Warehouse is done via a simple alter command. Snowflake emphasizes that this process requires no downtime. Customers can convert their production workloads in batches, while retaining existing warehouse names, policies, and permissions. Below is an image illustrating the differences between creating a standard warehouse and Adaptive Warehouses.
The approach addresses a common complaint that Snowflake identified among users. Warehouse names are often hardcoded in pipelines and scripts, making it a time-consuming and disruptive process to consolidate workloads. With the new service, organizations can optimize their infrastructure without operational disruption.
Better price-performance ratio
Adaptive Compute leverages the latest hardware and performance enhancements available from Snowflake. The service is designed to improve performance through regular refinements to the core engine automatically.
Optimized resource sharing ensures that queries can benefit from a shared pool of clusters within the account, allowing them to leverage a more efficient resource allocation. This should increase efficiency and reduce costs for organizations running multiple workloads.
Adaptive Compute is currently in private preview. Snowflake has not yet announced a definitive release date for general availability. The service is expected to become part of the broader offering to help organizations get more value from their data without the complexity of infrastructure management.
Standard Warehouse Generation 2 brings significant performance improvements
In addition to Adaptive Compute, the Snowflake platform offers several new features. The cloud data provider has focused on performance optimizations in addition to automation. With Standard Warehouse Generation 2 (Gen2), Snowflake promises significantly faster query execution. This improved version includes updated hardware and additional performance enhancements. Snowflake claims that customers have experienced 2.1x faster performance for core analytics workloads over the past twelve months.
Tip: Snowflake Openflow extracts data from any business system