Turbonomic now stitches Azure App Service plans and web apps into the supply chain and automates plan rightsizing and wasted plan remediation. 

Background  

IBM Turbonomic discovers and analyzes resource utilization trends of cloud infrastructure to automate optimization actions. Optimization is achieved by providing the resources required to serve application demand while minimizing costs and staying compliant with business rules. Turbonomic discovers cloud-hosted virtual machines, volumes, databases, database servers, and containers, and monitors utilization of CPU, memory, disk I/O, IOPS, throughput, storage, and other performance metrics to strategically reconfigure capacity. Now, Azure PaaS customers can realize additional cost savings and reduction of management overhead by using Turbonomic to automate App Service plan rightsizing and eliminate waste.   

Enhancement 

Turbonomic Azure customers have realized platform-as-a-service (PaaS) optimization opportunities for years. By first automating rightsizing of Azure SQL DTU Databases, Turbonomic demonstrated the ability to optimize PaaS entities. After that, optimization actions for vCore databases were made available. Now, cloud-hosted applications running on pure PaaS can be optimized via App Service plan rightsizing and waste elimination.  

First, Turbonomic automates deletion of App Service plans with no associated applications. Taken alone, this capability has the potential to recover 7-28% of dollars spent on App Service and reduce the number of App Service plans by 8-19%. After eliminating unused plans, an additional 20-40% of dollars spent on App Service can be saved by taking Turbonomic’s App Service plan vertical scaling actions.  

The Model 

Generalized representations of App Service entities are used in the Turbonomic supply chain and other platform views. Plans are modeled as Virtual Machine Specifications (Virtual Machine Specs), and web apps as Application Component Specifications (App Component Specs). Since a single App Service plan encompasses one or more virtual machine instances with the same memory, CPU, and storage capacities, it can be thought of as the specification for each underlying VM. A Virtual Machine Spec’s “Number of Replicas” commodity indicates the number of VM instances underlying the represented App Service plan. The same goes for web apps – Application Component Specifications are used to depict web apps since each app has one or more instances whose performance characteristics are unified into a single representation.  

Clean Up 

Empty App Service plans are those that have no associated running applications. They consume unnecessary resources, increase carbon emissions, and incur costs, but can be easily removed. While vertical scaling actions are generated on Standard, Dedicated, and Isolated plans only, delete actions are available on those types as well as Elastic Premium and Workflow Standard plans. To prevent deletion of plans that have been recently provisioned or are in the process of being repurposed, delete actions can be enabled on plans that are empty for a specified duration. Turbonomic automatically detects and deletes empty App Service plans, reducing wasted resources with no impact on application availability or performance. 

A few key pieces of contextual information are available in Action Center and expanded delete action detail views. While App Service plan deletion actions are irreversible, they are also non-disruptive because no applications are drawing resources from plans identified for deletion. Since deleted plans are removed entirely, the total plan cost is saved. The action detail view shows the number of days a given App Service plan has been empty and any applied user-defined policies as shown below.  

Rightsizing 

Dynamic vertical scaling of App Service plans is another key feature of Turbonomic’s App Service optimization solution. The platform analyzes historical resource utilization data to determine if an App Service plan is over- or under-provisioned. Vertical scaling actions are recommended by leveraging percentile calculations of memory, CPU, and storage utilization, and executed by changing plan pricing tiers. If an App Service plan is over-provisioned, IBM Turbonomic can dynamically scale it down to reduce costs. If an App Service plan is under-provisioned, IBM Turbonomic can scale it up to meet the demands of applications running on it.  

Turbonomic’s App Service plan vertical scaling actions are aware of constraints imposed by Azure. Actions respect instance type availability constrained by region, server rack, deployment slot and hybrid connection utilization, zone redundancy, Hyper-V enablement, and isolation type. Turbonomic takes the guesswork out of App Service plan sizing by considering instance type restrictions imposed by Azure combined with those implemented by users in the form of policies. By updating tier availability for all plan instances on a regular cadence, the system only considers currently available configurations when making optimization decisions.  

The Action Center and action detail views illustrate impacts on resource utilization and cost while providing other contextual information about actions. Historical resource utilization trends are visualized in charts on the left side of the action detail view and displayed textually in the Resource Impact grid. The Cost Impact grid shows the cost of a single plan instance and the number of instances currently allocated to the plan. Total cost is calculated by multiplying the instance type cost by the number of instances, and the total savings/investments values represent the difference between current and projected plan costs. The action detail view also specifies that vertical scaling actions are disruptive but reversible, and represents any constraints imposed by either Azure or user-defined policies.  

Conclusion  

IBM Turbonomic provides an effective solution for optimizing Azure App Service deployments. By removing empty App Service plans that are not in use and dynamically scaling plans based on historical resource utilization, Turbonomic can help organizations improve performance and lower costs while adhering to business guidelines. Tell us about your experience with IBM Turbonomic’s Azure App Service optimization capability, and how we can help you further reduce cloud costs.