Anderson Oliveira
Senior FinOps Lead at AXA and Timestamp Portugal
By Anderson Oliveira
Addressing multidimensional cost showback requires a bit of a change of mindset. Rather than thinking of cloud costs as a single object, we see it through three dimensions:
All of these ‘Views’ have different ownership within the organization and the total cost of cloud tracking must be looked at together.
The tooling suite used to manage these different views has been internally developed, enabled by ingesting billing data from CSPs. The variable criteria used to determine the cost splits are based on Prometheus data. This data is collated and managed using Airflow, Big Query and Looker. The system relies on daily pulls of resource data to provide insight across all dimensions. As the company is employing a multi-cloud strategy, consolidating all the data in one place and using a common set of tags across each provider is paramount for success.
The Resource View allocation uses a simple direct tag representing the owning organization. The Technology platforms have responsibility for the direct resources (Compute, Storage, etc.). They determine the capacity of resources needed and make decisions on the appropriate type for platform requirements. They are responsible for making decisions on commitment levels, duration, and right sizing potential to ensure the resiliency of their platforms.
Microservice View owners share the cost of multiple cloud resources, to provide specific capabilities required by the Tenants. The Microservice View allocation is a proportional share of the Resource costs consumed using a methodology based on variable usage (resource share x usage metric, such as time used, for example). The Microservice View relies on unique tags, such as namespaces within Kubernetes clusters, to determine the share a resource is used.
The Tenant View (a.k.a., the business view) represents the cost share of total hosting for allocation to internal products (customers or B2B). The current allocation methodology is a fixed proportional split, based on unit economics. The team is exploring using an operational metric to pivot towards a more real-time variable split, such as using API calls into each Microservice.
In the realm of cloud budget management, it is important to note that while there exists a singular budget for cloud-related expenditures, the responsibility for overseeing this budget is collectively distributed among various perspectives. Broadly speaking, each of these perspectives undertakes a thorough examination of the distinct business requirements to ensure the provision of diverse allocation methodologies that facilitate crucial business decision-making.
At a more strategic level, these methodologies take into account the following key considerations:
By adopting these multiple viewpoints and methodologies, we aim to ensure a comprehensive and well-informed approach to managing our cloud budget, aligning our expenditures with the specific needs and priorities of the business.