In the second part of this blog series, I will summarize the key differences between my datacenter and the public cloud as part of FinOps implementations for data platforms. Although similarities can be found in many areas, many of the differences stem from fundamental differences in cost structures, resource allocations and operating models between the two environments. The situation can be best illustrated by comparing the data platforms from a FinOps perspective:
Cost model and service pricing
The differences in the cost model are clear and best explained by the classic Capex / Opex approach:
| Public cloud | Dedicated data centre |
|---|---|
| Pay-as-You-Go: Costs are based on current resource utilisation and billed monthly, providing detailed visibility into resource consumption (e.g. computing power, data movement and storage). | Fixed costs: Costs are fixed in nature and can be amortised through various payment schemes. Large up-front investments are made in infrastructure (hardware, licences, premises). The running costs of a data centre (e.g. cooling and electricity) are recurrent in nature. |
| Flexible pricing: customers can optimise their costs by committing to specific services and capacities, thereby achieving savings. Public Cloud offers customers an agile way to consume according to usage and avoid up-front investments. With agility may come higher costs and these should be analysed together with the business benefits. | Static pricing: the cost is constant regardless of usage. Costs can be allocated according to users, but the total cost to the organisation is the starting point for allocation, thus limiting the pricing possibilities within the organisation. An organisation can achieve cost efficiency through various consolidations and optimisations to reduce the total cost to users. |
Scalability of resources
Resource scalability and flexibility are important factors in an organisation’s FinOps matrix, which aims to optimise the cost of the resources required. Optimisation typically involves reconciling the flexibility of the cloud with the cost-effectiveness and performance of your own data centre. Put simply, the cloud is a good solution for needs where resource requirements are difficult to predict, and a dedicated data centre often justifies its place when resource requirements are well known.
| Public cloud | Dedicated data centre |
|---|---|
| Scalability on demand: resources can be scaled up or down dynamically based on workload. There is virtually no limit to the resources available in the public cloud. | Limited scalability: scaling up may at some point require the purchase, installation and configuration of new hardware, which takes time. Downscaling may be difficult. |
| Support for elastic workloads: ideal for variable workloads such as batch runs or machine learning, as well as highly seasonal events and the introduction of new services. | Capacity constraints: the physical capacity of the data centre limits performance and scalability and can lead to situations where resources run out without further investment. |
Cost allocation and accountability
Cost allocation is easier in the cloud. On the other hand, you may have to pay more for the same capacity than in your own data centre. In-house data centre costs are charged to the organisation on a 24/7 basis, while services purchased from the cloud are priced according to usage. For example, if a customer’s activity is concentrated on eight hours a day, why pay for an extra 16 hours?
| Public cloud | Dedicated data centre |
|---|---|
| Detailed cost allocation: costs can be allocated to specific projects, groups or applications through different tags, accounts or subscriptions. | Rough cost allocation: Costs are less detailed and are often allocated on the basis of estimates or fixed percentages, unless detailed usage monitoring is carried out. However, the total amount of costs, i.e. the cost to be allocated, always remains the same. |
| Showback / Chargeback. | Customized Showback/Chargeback: Implementing a comprehensive model requires manual allocation or procurement of third-party solutions. |
| Public cloud | Dedicated data centre |
|---|---|
| Opportunities for cost optimisation: monitoring is easy, but cost optimisation requires familiarity and the same depth of knowledge as for in-house data centres. For cloud computing, the opportunities for cost optimisation are significant and impactful, as optimisation can quickly have a visible impact on an organisation’s overall costs. | Opportunities for cost optimisation: optimising the costs of in-house data centres is more about optimising the optimal and efficient use of resources than about reducing costs. In other words, more can be done with the same amount of resources. This is because the investments made have determined the minimum level of total costs. It would therefore be useful to compare in-house data centres and public clouds on the same table in order to objectively decide on the most cost-effective way of providing the necessary services. |
| Service-specific cost data to reduce overall costs: near-instant access to usage data allows proactive cost management in the short term on a service-by-service basis. The organisation pays only for the resources it uses. | Service-specific cost data for the correct granularity of total costs: usage data allows for the granularity of costs, but does not allow for a rapid reduction of the total cost level to be granularised. However, this ensures a fair allocation of costs. |
Optimized strategy
There are many similarities between the optimisation possibilities of data platforms in public cloud and in-house data centres, although in-house data centres offer a wider range of workload consolidation possibilities and typically much greater improvements in cost-effectiveness than in public clouds.
| Public cloud | Dedicated data centre |
|---|---|
| Workload optimization (Right-Sizing) and consolidation: workloads can be consolidated at the database or database instance level into larger Managed Instances or servers in the cloud to optimize hardware usage. In addition, the size of virtual machines and storage tiers can be optimised on a per-use basis to save costs. | Workload optimization (Right-Sizing) and consolidation: workloads can be consolidated at the database, database instance and virtual machine level into larger servers to optimize hardware usage. |
| Reserved instances and savings plans: clouds offer the possibility of the lowest prices if you are willing to commit a certain amount of resources.In this case, you have to compromise on the fully consumption-based billing of the cloud. | Lifecycle management: consolidating and optimising workloads to ensure that the equipment in use delivers optimal value. Older investments that are obsolete and of limited added value can be replaced by the most cost-effective solutions. Extending the life cycle of equipment is also an option. |
| Dynamic resource scaling: public clouds have the possibility to automatically scale resources according to usage needs. | Workload timing: this optimises the timing of load execution to reduce power and cooling costs. Timing the execution of workloads ensures efficient use of resources and also cuts down on load spikes. This eliminates, among other things, over-investment and unnecessary use of electricity and cooling. |
Automation
In public clouds, automation is built-in, while in dedicated data centres it has to be built using separate automation solutions. Both can achieve a very high level of automation, but the means and ways of achieving it are different.
| Public cloud | Dedicated data centre |
|---|---|
| Extensive automation options: possibility to use serverless technology, scripts and orchestration tools for dynamic scaling and resource management | Manual model or limited automation: automation may require significant investment in orchestration tools (e.g. Ansible, Puppet or Kubernetes, PowerShell, Dbatools) |
| Infrastructure as Code (IaC): service deployment and scaling can be automated using IaC tools (e.g. Terraform, Bicep). | Hardware configurations: in-house data centres may have very different hardware configurations that are not designed to act as a large common resource pool. Virtualization helps, but hardware-specific features still affect optimization and consolidation. SQL Governor software has various features that make it much easier to compare and scenario different configurations in terms of capacity needs and performance requirements. |
Economic flexibility
A public cloud is a more economically flexible model than a dedicated data centre. There is a legitimate place for both the cloud and the dedicated data centre, but optimising their roles requires a deeper understanding and calculation. A categorical position on the advantage or disadvantage of one or the other does not stand up to further analysis.
| Public cloud | Dedicated data centre |
|---|---|
| High financial flexibility: all costs are operating expenses (OPEX) and organisations can avoid capital and investment costs. | High cost of capital (CAPEX): requires upfront capital and investment and this can make it difficult to adapt to changing economic conditions and market changes. |
| Burst capacity and scalability: The public cloud is able to cope with short-term and undiversified workloads without large investments. | Capacity constrained: Burst management requires overprovisioning or leasing of excess capacity, both of which are typically economically costly options |
Governance and compliance
In terms of governance and compliance (regulations and standards), the cloud is clearly a stronger and more cost-effective option than a dedicated data centre. This is particularly true in highly regulated environments. For public clouds, it is good to consider the location of your data platform and data. Many clouds may allow access to data globally and it may be difficult, or even impossible, to pinpoint the location of the data completely.
| Public cloud | Dedicated data centre |
|---|---|
| Policy enforcement: for example, tools such as Azure Policy or AWS Config enforce usage and cost management across user accounts, which contributes to greater management transparency. | Need for custom policies: good governance requirements typically need to be implemented by third-party tools. |
| Built-in security and compliance: public clouds provide compliance with global standards (e.g. GDPR, HIPAA). | Custom compliance measures: organisations need to ensure that physical and functional compliance requirements are met. |
| Public cloud | Dedicated data centre |
|---|---|
| Collaboration between teams: financial, IT and various planning teams can access information relevant to them through views optimised for them. | Fragmented functions: collaboration between different functions may be less integrated and result in slower operations. The introduction of new services can be slow and even disruptive to business. Machine automation can be used to implement cloud-like automation, but this means additional investment. |
| Built-in security and compliance. Automation helps to eliminate obvious user errors. | Centralised management: typically IT manages the preparation of resources, which can lead to potential bottlenecks and slowness in deploying services. |
Closing words
The differences between the public cloud and your own data centre can be summarised in the table below:
| Public Cloud | Private Data Center | |
|---|---|---|
| Cost Flexibility: | Pay-as-you-go, elastic pricing | Fixed costs, static pricing |
| Scalability: | On-demand, highly elastic | Limited, capacity-constrained |
| Allocation: | Granular, usage-based | Coarse, manual tracking required |
| Automation: | Extensive, cloud-native tools | Limited, custom solutions needed |
| Financial model: | Operational Expense (OPEX) | Capital Expense (CAPEX) |
| Visibility: | Real-time, granular insights | Delayed, manual aggregation |
While both environments benefit from FinOps practices, the public cloud emphasises flexibility, transparency and dynamic cost management, while dedicated data centres focus on fixed cost optimisation, resource utilisation and capacity planning. Implementation strategies and tools differ significantly due to underlying architectural and financial differences.
Optimal solutions are often found at the interface between these two worlds, in so-called hybrid solutions. There is a place for both, and finding the most cost-effective solution requires a thorough understanding of both and how to make the best of them.
The positive user experience of public clouds has motivated solution providers in their own data centres to focus on developing a cloud-like user experience. It is good to remember that the cloud is not a place but a way of doing things. With DB Pro’s SQL Governor solution, you can see the status of both your own data stores and the public cloud, and make effective FinOps decisions for performance and capacity optimization of your data platform in a controlled and centralized manner. You can read more about SQL Governor’s features here: www.sqlgovernor.com
Jani K. Savolainen
jani.savolainen@dbproservices.fi
0440353637
VP & Chairman
