twobee - Fotolia
As organizations migrate more workloads to the cloud and adopt a hybrid or multi-cloud strategy, they also face concerns about vendor lock-in. One way to address this issue is to practice cloud arbitrage.
In this model, IT teams regularly compare vendor pricing, performance and overall capabilities, then move workloads to the platform that best meets those needs, with the ultimate goal of saving money. Developers and administrators need to build and manage cloud-agnostic applications for this to succeed. And workload migrations can still be a challenge, depending on the amount dependencies, the differences between clouds and possible egress fees.
However, as budgets tighten, more organizations should consider adopting cloud arbitrage as a valuable tool to keep cloud spending in check.
Cloud arbitrage tool suggestions
As of publication, there's no managed cloud arbitrage tool offering. Customers need to put their own tools in place to have total visibility into their cloud management, which can then be used to support a cloud arbitrage model.
Quite often, this starts with having a cloud management platform that includes a service catalog and the cloud economics tools to track and analyze your cloud spending for trends. Another component for cloud arbitrage is an infrastructure management option that supports provisioning and management across clouds.
For example, you could deploy HashiCorp's Terraform and Nomad orchestration tools to drive multi-cloud provisioning. Hence, the organization always chooses the lowest cost provider and instances.
Cloud arbitrage usually relies on containers. Applications and their dependencies can be packaged inside a container and moved to another environment much more easily than with VMs alone. Expect to see container orchestration platforms like Kubernetes play a major role in cloud arbitrage going forward.
Cloud arbitrage best practices
Organizations that look to implement cloud arbitrage want to deploy workloads on the most cost-effective platform. Yet, in today's highly competitive cloud market, the prices for primary cloud services don't change enough for most offerings to make cloud arbitrage worthwhile.
The most significant savings and pricing fluctuations in the cloud market are for VMs provisioned from excess capacity, such as AWS Spot Instances, Google Cloud Preemptible VM instances and Azure Low Priority Virtual Machines or Spot VMs. There are still discounts available for up to 90% when compared to on-demand instance pricing.
Spot instances are often used for batch analytics jobs or as part of automatic scaling to support spikes in traffic. These VMs can be shut down with a few minutes notice if the provider's system requires the capacity, so they should only be used if your application can tolerate the disruption. If you plan to use these discounted VMs in your cloud arbitrage, make sure your application is built with these characteristics in mind.
While cloud arbitrage is usually associated with multi-cloud, customers don't need to move to an entirely different cloud provider to gain cost savings. Upgrading to a newer instance class of virtual machine can save 15% with minimal downtime or impact to operations.
Alternatively, buying reserved instances for specific use cases, such as workloads that must be on 100% of the time, can save 25% or more for users. This option makes sense for organizations that have the data to ensure they're reserving the right instance class and size. However, you must know the breadth of each cloud provider's offerings -- and how best use them -- for the highest return on investment.