The migration to cloud is obvious. Organizations are adopting cloud computing for all variety of applications and use cases. Managed cloud services, commonly referred to as software as a service (SaaS), offer many benefits to organizations including significantly reduced labor costs for system administration and maintenance, as many of these costs are shifted to the software vendor. SaaS also provides organizations with faster time to value as they adopt new technologies by eliminating the need to acquire and configure hardware, and it also eliminates the need to install software. In fact, we assert that by 2025, nine in 10 organizations will be using multiple cloud applications in order to minimize the costs of administration and maintenance. Yet, there are some challenges associated with cloud computing I’d like to address in a series of Analyst Perspectives:
The leap to cloud computing has been met with some surprising operational realities. For instance, the cloud introduced new pricing models and cost structures that are different from traditional on-premises enterprise deployments. Many cloud pricing models convert fixed costs to variable costs, with hardware costs often built in or hidden. Traditionally free data access has been replaced by vendor ingress and egress charges to move data in and out of a cloud ecosystem. These charges alter the economics of data processing. Different pricing models are not necessarily bad, but IT organizations previously understood and could project on-premises software and hardware costs.
I like to compare cloud cost models to different pricing models for transportation via automobile. You can buy a car and pay for all the related costs such as maintenance, insurance, fuel and tolls. You can lease a car for a designated period of time, and you would still be responsible for most of the related costs. These two would be analogous to on-premises, self-managed deployments.
Alternatively, you can rent a car, which you drive yourself, typically for shorter periods of time, with little or no responsibility for any additional costs other than fuel and tolls. And, for very short-term requirements, you can engage a ride-sharing or taxi service which includes a driver. These would be analogous to cloud pay-as-you-go models.
If you have kids and you allow them to use your credit card for their ride-sharing activities, you can see how the pay-as-you-go cloud model could easily go awry. Within organizations, often the purchaser of the cloud services doesn’t have direct responsibility for the cost of the cloud services. They may be making purchases that will be expensed — perhaps in the early days of exploring or consuming a new cloud service — or there may be a centralized purchasing contract covering multiple departments with no tracking to determine how those costs should be allocated.
Many organizations have experienced constantly expanding workloads, regardless of whether they are deployed in the cloud or not. The volume, variety and velocity of data required to support an organization’s digital transformation efforts continues to grow, with unstructured, streaming and external data sources now standard components of information architectures. Artificial intelligence and machine learning (AI/ML) models that extract insights from modern data stores require high-speed parallel processing, creating compute-intensive workloads. And delivering on the expectation of lower latency and improved responsiveness to internal and external customers requires much more frequent processing of data. All of these forces tend to increase computing costs over time.
One way to mitigate (cloud) cost surprises is by monitoring cloud services which provides visibility into patterns of usage and performance. Consoles and reports offer an easy solution for highlighting unusual usage patterns, preventing situations where usage is not measured or measurements are not reported in a timely fashion. Alerts and thresholds can compare with baselines to identify runaway processes and raise awareness of potentially unauthorized use or high-use scenarios that may be inflating costs.
When considering costs of cloud computing services, it’s important to consider not only subscription costs but total cost of ownership. This includes the cost of resources to manage, maintain, adapt and integrate the services into your business processes. Total cost also includes the cost to customize the software to meet your requirements. Sometimes, the options to adapt, integrate or customize may be more limited than on-premises software products.
Organizations must also master the variability of pricing models for cloud services. One size does not fit all when it comes to pricing. Pay-as-you-go, consumption-based pricing models have no upfront commitment and are best for experimentation and variable workloads. However, for highly utilized systems, pricing models based on a longer or larger commitment typically offer lower costs. In all cases, visibility into pricing and the ability for departmental charge-backs and allocations help align the organization’s behavior with the costs being incurred.
The objective of this perspective is not to discourage you from considering cloud-based services. Rather, it is to make sure you are aware of some of the issues to consider as you evaluate cloud computing options so you can go in with your eyes wide open. The next perspective will address the realities of hybrid and multi-cloud configurations.
Regards,
David Menninger