Turbonomic

Azure Synapse Analytics Dedicated SQL pools are now optimizable in IBM Turbonomic via automated idle instance identification and suspension. 

Background

Turbonomic has positioned itself as the industry leader in cloud optimization by automating rightsizing and clean-up of cloud infrastructure and platform-as-a-service (PaaS) resources. Particularly in the realm of PaaS, the platform’s capabilities surpass those of its competitors in both breadth and depth.  

Turbonomic first established competency in PaaS optimization by targeting relational database offerings. Instances of Amazon Relational Database Service (RDS) and both DTU and vCore purchase models of Azure SQL Database have been optimizable for years. Optimization of Azure App Service plans through rightsizing and wasted plan remediation was added earlier this year. Here, Turbonomic further expands its PaaS reach into the Azure service portfolio by offering its first data warehouse optimization.   

Enhancement  

Azure Synapse Analytics is an enterprise analytics service that provides data warehousing, big data analytics, and reporting capabilities to some of the world’s largest organizations. Users can store and query relational and non-relational data at an incredible scale, secured by the latest Microsoft technologies employing the highest standards. However, those capabilities don’t come cheap. The data warehouse and analytic compute resources together under the Synapse umbrella are collectively referred to as Dedicated SQL pools, whose instance types range in price from $876/month (DW100c) to $262,800/month (DW30000c) in the US East 2 region.  

Enter Azure Synapse Analytics Dedicated SQL pool optimization. By monitoring Dedicated SQL pool connection activity, Turbonomic identifies and automates idle instance pausing after a configurable duration. Customers have realized the ability to eliminate nearly 20% of Dedicated SQL pool costs, with additional point-in-time cost reductions of over 53%.  

The Model  

Synapse Dedicated SQL pools are modeled as Databases connected to regions in the Turbonomic supply chain. Dedicated SQL pools are differentiated from Azure SQL Databases by Pricing Model; Synapse entities can be viewed in isolation by applying a filter on Databases where Pricing Model is DWU.  Database Warehouse Units (DWU) and Storage Amount capacity and utilization are tracked for observability, while Connections utilization is monitored for suspension eligibility. 

 
 
Suspension   

Dedicated SQL pools are considered unused when connections aren’t made and there are no active queries on an instance. Though compute resources are allocated and available, workloads are not running but costs are still incurred. The separation of compute and storage resources allows for each to be scaled and managed independently.  

This separation allows for pool compute resources to be paused. When compute is paused, resources are released from the associated account and returned to the Azure data center resource pool. Compute costs go to zero. Storage remains intact along with all accompanying data and is charged accordingly, though storage costs account for less than 1% of instance cost in many cases. Compute can be resumed when use of the pool is required. 

By default, Turbonomic recommends suspending Dedicated SQL pools when no activity has been observed for at least one hour. This duration can be adjusted with ease through a dynamic group of databases defined by the Hours Idle filter. Customizing the duration before which suspension is recommended allows organizations to comply with business rules. 

Action Center and expanded action detail views provide useful context related to Database suspension recommendations. These actions are reversible since instances can be resumed with a single button-click or API invocation, and non-disruptive since Turbonomic checks instance activity before executing. No action is taken if activity has resumed since the suspension recommendation was made. The Idle Time duration is also made visible so users have a clear perspective of when the instance was last used. Finally, elimination of compute costs associated with the instance is made clear.

Conclusion

IBM Turbonomic identifies and automates the optimization of Azure Synapse Analytics. Synapse costs are reduced while performance is maintained by suspending unused Dedicated SQL pool compute nodes. Tell us about your experience with IBM Turbonomic’s Azure Synapse Analytics optimization offering and how we can continue to help you reduce cloud costs, or learn more about cloud cost optimization with Turbonomic here.