Multicloud is an emerging approach for deploying enterprise computing assets. Optimizing the costs of a full multicloud deployment can prove tricky, because there may no unified management plane for deploying, monitoring, and governing all distributed IT assets.
Multicloud cost models are shifting because best practices for deploying these distributed environments are in flux. The following trends are shifting the multicloud cost model:
- More cloud domains, more cost: Enterprises are growing the complexity of their cloud deployments, moving beyond hybrid public/private clouds into environments where they adopt two or more public clouds for various apps and workloads, while also establishing mesh and edge clouds that connect additional domains of infrastructure, platform, and application assets. Clearly, this growing complexity adds to the costs of storage, compute, and other hardware resources, as well as the associated software and IT personnel.
- More cloud layers, more cost: Users are hybridizing diverse virtualization, container, and serverless platforms over their multiclouds. This requires abstraction layers and the associated control planes to monitor it all. It also creates the need for driving real-time continuous integration and continuous deployment of cloud-native microservices from end to end. The costs of provisioning and managing all of these service and management layers can be considerable.
- More cloud workloads, more cost: As organizations modernize their IT environments, they are refactoring, containerizing, and orchestrating more functionality as cloud-native microservices. The cost of modernization, as well as of migrating legacy assets and workloads to the cloud, also add to the cost overhead of going fully multicloud.
In the face of these trends, enterprise IT should heed the following guidance to keep multicloud costs under control:
- Provide a full accounting of multicloud costs to stakeholders: Multiclouds can become a big overhead item if you don’t watch out. Enterprise IT should adopt a comprehensive multicloud cost tracking and accounting system to keep tabs on total cost of ownership. Per gigabyte storage costs are a key piece of multicloud TCO, but so are labor, power, disaster recovery, and square footage. The cost accounting should provide multicloud users with visibility into their usage of compute, storage, and other IT resources across the multicloud. It should also deliver detailed cost reports that flag trends and anomalies in usage of multicloud resources while recommending remedial strategies for avoiding resource overprovisioning.
- Avoid proliferating superfluous clouds: Multiclouds involve two or more internetworked clouds that address a common use case, such as an end-to-end data science pipeline. According to a recent RightScale study, 70 percent of enterprises are embracing hybrid cloud by deploying multiple public and private clouds, with the average cloud-using organization running applications in four clouds, and experimenting with four others. However, cost control may require that enterprises avoid going the multicloud route when a single public or private cloud will suffice. For example, your incumbent public cloud provider may offer the most cost-effective capabilities for data ingest, preparation, modeling, training, and inferencing, thereby weakening the case for running any of these workloads on another public cloud or even on-premises in your data center.
- Keep tight reins on multicloud capacity: Multicloud cost control depends on rightsizing the compute, storage, memory, and bandwidth resources you’ll need for expected workloads, while building in the elasticity needed to absorb growth. Eliminating waste in multicloud provisioning involves deduplicating data, monitoring utilization, provisioning rightsized instances and storage, automating shutdown of idle or underutilized workloads and services so that they don’t consume resources indefinitely, and shutting down low-priority workloads during peak hours. It also involves making sure not to overprovision expensive edge-based storage and compute when more cost-effective resources are best deployed in bulk in the cloud.
- Automate movement of workloads to data sources on demand: Multiclouds might become a huge bandwidth hog if you’re constantly moving large amounts of data to centralized servers for processing. This is especially true as mesh multiclouds become more common and the data exchanges amongst them become more frequent. Cost control in the multicloud will require the automated movement of workloads so that they execute at data sources, as well as automated applicaton of security, governance, and other policies to these workloads and data. One important use case in this regard will be automated movement of centrally created and trained machine learning models to edge nodes for real-time inferencing. In this regard, cost efficiencies may demand that more processing be done at the edge, where data is increasingly being captured and persisted and where algorithmic decisions are being automated.
- Arbitrage regularly across alternative clouds to get the lowest cost: Multiclouds can be difficult to optimize from a cost standpoint if you lack visibility into what different public cloud charges for the same workload—such as online backup and disaster recovery—or are unable to easily move workloads and data across clouds to take advantage of cheaper rates. To keep multicloud costs in check, you should consider using tooling or even a managed service provider to identify the best rates on various public clouds and also to automate movement of workloads and data to take advantage of them. The better managed service providers may help you to do this arbitraging on an application-by-application basis to achieve more granular multicloud efficiencies, perhaps even identifying when it’s most cost-effective to keep cloud workloads on-premises.
As enterprise IT professionals get their arms around multicloud cost optimization, they’ll want to explore the growing range of commercial solutions for that. Over the past years, several leading IT vendors have released multicloud management tools, though strong end-to-end cost tracking and management capabilities are only beginning to enter this segment. Here are the most noteworthy multicloud management tools:
- IBM Services for Multicloud Management: IBM’s new IBM Services for Multicloud Managementsolution supports self-service acquisition and management of IT resources across multiple cloud providers, on-premises environments, private clouds, legacy infrastructure and container environments. It automates the deployment of services of different types and from different vendors to be integrated easily and made available to consumers. It also supports monitoring and maintenance of systems, including legacy infrastructure, private cloud, public cloud and container environments. And it integrates with the ServiceNow Portal for centralized purchasing, configuration, monitoring, and cost tracking and management of cloud services and solutions from multiple providers.
- Google Cloud Anthos: Google Cloud’s new Anthossolution lets enterprises centralize management of applications in the multicloud. It integrates its customers’ on-premises resources with compute, storage, data and other resources in its public cloud, as well as with equivalent resources in competitors’ public clouds. Anthos is an open distributed infrastructure for spanning multiclouds. It enables cloud-native applications to run unmodified on the hardware platform of the customer’s choice, and to run an app anywhere unmodified, either on existing on-premises hardware platforms or in various public clouds, not limited to Google Cloud Platform. It also automates deployment guardrails, role-based access controls, configuration and capacity management, and other policies at scale across cloud deployments, including Istio services meshes. But it lacks a multicloud cost tracking, management, and reporting capability.
- Cisco CloudCenter Suite Cost Optimizer: Cisco’s newly enhanced CloudCenter Suite simplifies design, deployment and optimization of applications and infrastructure across multiple cloud environments. Available on a software-as-a-service subscription basis, one of its new modules is Cost Optimizer. This enables users to analyze their consumption patterns related to usage of private and public cloud resources. It provides visibility into their total cloud spending for compute, storage, network and other services. It also identifies cost-optimization strategies to help customers minimize cloud overprovisioning and optimize the sizing of their cloud workload instances. The module’s cost-reporting capability is available in all of the suite’s licensing tiers, while the add-on remediation and recommendation capabilities are only available in the Advantage and Premiere tiers.
From the recent Cisco Live EU 2019, here’s a good discussion of Cost Optimizer by Dave Cope, senior director, market development, Cisco Cloud platform.
