Cloud deployments offer opportunities for IT to gain performance and reduce maintenance tasks—cloud rightsizing is the way to grab these opportunities and lower cloud bills. But the ease of initiating cloud instances coupled with automated instantiations on a needed basis lead to a false assumption that deployments are running in their most appropriate configurations.
They may be operating just fine, but that doesn’t mean they are optimized for cost. IT managers overseeing cloud-based environments need to look closely at the details of their systems because they can develop bad behaviors that may not be immediately evident—but they can show up later in cases of cloud systems that translate to unnecessary charges and bloated invoices.
Here are the six most important factors CIOs need to review as they right-size cloud-based infrastructure and continue their efforts in cloud optimization.
Whether you go with AWS EC2, Azure VMs or GCP Compute Engine, the building blocks of cloud instances are compute resources offered as virtual machines. Each provider has its own nomenclature to describe the capacity of the instance, and these can range from relatively low capacity offerings with the equivalent of a single CPU with minimal memory to multi-CPU configurations with added GPU and large amounts of memory for more intensive compute tasks. An advantage to cloud computing environments is that configurations can be upgraded or downgraded on demand. Some will perform upgrades automatically based on usage and how the customer has instructed the service to adapt to changes. But that same flexibility that makes services run smoothly and quickly can lead to unexpected higher costs when usage conditions are not continuously monitored.
»Read more: Eliminate Cloud Sprawl
Over-provisioning happens when an instance is first created due to over-exuberant expectations or by overestimating the amount of processing required for a given load. New instances should be configured at relatively low capacities unless there is some history to indicate the need for higher performance. Automated capacity increases should be enabled to accommodate processing needs beyond original expectations, but that automation must be reviewed and monitored to assure instances are not automatically over provisioned.
Every developer thinks their app is a snowflake. It needs the most performant memory, the best processor speed, etc. However, not all workloads require tier 1 memory, so provision accordingly. Make informed architecture choices from the get-go to set yourself up for success.
Cloud instances may be charged by the second, minute or hour. There is a lot of granularity to choose from. This doesn’t have a lot of baring on environments and applications that need to be available continuously. But for development environments that work regular business hours, those idle 12 hours when devs are off the clock can add up. Schedule downtime for unused cloud resources to trim invoices. Check with the providers to determine whether they have automated facilities that will restart services if some enterprising developer decides to put in some extra time at night or during weekend hours. Turn off unused resources for Azure, AWS, and GCP cloud optimization.
Automated elastic expansion of cloud resources keeps your operations running at top performance. But being elastic doesn’t necessarily mean the resources will be compressed to a lower performance tier after the performance need has passed. CIOs should be mindful of the state of their compute environments and check for cloud rightsizing on their instances. Understanding what services are available for any given need can make a significant difference in monthly charges.
Modern cloud services offer a variety of automation tools that can be used to configure individual instances without manual intervention. While the tools have commonalities across the different cloud providers, they each have their own nuances and controls. IT can benefit by understanding what tools are available at each provider and implementing them to benefit their use cases. Investigate third-party tools that can manage rules and configurations across multiple platforms, making the process easier and more automated.
Cloud providers offer flexible services that can be easily configured to meet the needs of modern computing demands. CIOs, in consort with a cloud center of excellent (CCoE), have an opportunity to leverage these offerings by creating multi-cloud environments that take advantage of the best technology of each provider. But they need to be vigilant and understand what offerings deliver the best cost vs. performance for their needs--and which offer up appropriate cloud rightsizing. More than that, they need to evaluate the specific operational needs of each of their cloud-based applications and configure them or move them to new providers as conditions change. Reserved Instances (RI) save up to 70% over standard cloud billing for production workloads, but those savings are only realized if the RIs are used effectively.
IT takes advantage of the performance of commercial cloud service providers with vigilance on how every instance is allocated and billed. IT leaders make a significant difference in their expenses by monitoring their invoices against the actual needs of each application.