The digital landscape is rapidly evolving, and the question has progressed from whether to adopt cloud technology to how to effectively use it to drive business growth and innovation. Having recognized the power of the cloud in terms of agility, scalability and innovation, organizations are turning their focus to how to get more from their investment. One such consideration is multi-cloud.
Here, multi-cloud refers to the use of multiple cloud computing services from different providers, which can include both public and private clouds, inherently encompassing a ‘hybrid’ cloud approach. Studies show that when compared to single-cloud provider options, a multi-cloud environment allows choice among ‘best-of-breed’ services. Multi-cloud is the new norm for most enterprises.
In this article, we look at key drivers behind the rise of multi-cloud adoption, explore the main challenges enterprises face, such as cost management and security, while suggesting practical insights to help businesses build effective multi-cloud strategies.
Key Drivers of Multi-cloud Adoption
In this section I will cover a selection of the key drivers that are pressing people to move to a multi-cloud configuration
Flexibility
The global presence of major public cloud providers like AWS, Azure and Google Cloud allows organizations to select the most suitable provider for their workloads.
One of our customers was executing a large on-premises data center exit program, and they preferred AWS cloud for their workloads. They adopted a lift and shift strategy to fast track the program. With majority of their database workload being on Microsoft SQL Server, and to get the maximum benefits through Bring Your Own Licenses (BYOD) and Azure Hybrid Benefit (AHB) schemes, they opted to migrate their database workloads to Azure cloud while migrating rest of the compute workloads to AWS.
This strategy allowed them to leverage their existing Microsoft licenses in the cloud, resulting in a more cost-effective solution.
Avoid Vendor lock-ins
Relying on a single cloud provider can lead to risks such as service outages, increased security vulnerabilities, innovation constraints, and difficulties in meeting compliance requirements.
Cloud resources typically have planned maintenance windows. Consumers are informed about this well in advance via service alerts or notifications for them to be prepared for the downtime. In one instance, an Azure SQL Managed Instance was undergoing maintenance in a specific Azure region, with expectations that it would not affect the dependent workloads. However, it turned out to be a severe outage for the customer workloads that are heavily dependent on this service. Problems started with frequent database connectivity issues caused by unexpected network packet drops. Though a disaster recovery environment was available, it was a risk to trigger a failover as the data could be in an inconsistent state. With no alternate environment to recover from, this had an impact on the end users till the problem was completely resolved by the cloud provider.
Cloud-native tools like Azure VM Backup, AWS Step Functions, and Azure Logic Apps are tightly integrated with the provider’s ecosystem, which can lead to vendor lock-in. One customer had to move away from Logic Apps based implementation due to high recurring costs. Since the customer was already spending on the on-premises infrastructure for their workloads, they decided to host the solution on-premises with no added infra cost incurring to them. However, this was not an easy move, as the Logic App workflow is not portable/deployable as an on-premises solution, resulting in a complete re-write of the business logic into a .Net core solution.
Growing Adoption of SaaS and PaaS
The rise of Software-as-a-Service (SaaS) and Platform-as-a-Service (PaaS) solutions is encouraging multi-cloud strategies, as many organizations use cloud-native services from different providers. For example, many organizations use Salesforce’s Customer Relationship Management (CRM) software, offered as SaaS, in conjunction with other cloud providers to build custom applications (via PaaS offerings) that seamlessly integrate with their existing systems.
In adopting a multi-cloud SaaS strategy, a customer chose to host their workloads on AWS for scalability and performance, while using Power BI service on Azure for advanced reporting and analytics. This decision was primarily driven by their users’ familiarity with the Power BI interface, ensuring a smooth transition and minimizing the learning curve.
AI, ML and Analytics
Organizations are using multiple clouds to exploit the strengths of different providers in AI, machine learning, and analytics. Google Cloud is strong in data analytics with BigQuery, while Azure is popular for enterprise AI and machine learning.
For example, we had a customer in the energy management domain that deals with IoT generating a lot of data that need to be classified further. The manual classification approach was not scalable, and they approached this as a typical Machine Learning problem of data classification. Though their workloads are running on AWS cloud, they explored Azure AI/ML capabilities and adopted Azure services for their enterprise AI and machine learning.
By using AI and ML services across multiple clouds, companies can accelerate innovation and competitive advantage.
Regulatory compliance
Regulatory compliance, especially for data sovereignty and industry-specific requirements, is driving demand for multi-cloud deployments in financial services, healthcare, and government sectors. For example, organizations in Australia are adopting a multi-cloud strategy to ensure a credible business continuity plan to comply with APRA CPS 230.
To comply with the regulations, a banking customer in Australia set up their secondary environment in Azure cloud while AWS remained as their primary cloud for their data heavy applications.
Disaster Recovery (DR) and Business Continuity
As mentioned above, many organizations adopt multi-cloud strategies to improve disaster recovery capabilities. By distributing workloads across multiple cloud environments, businesses can ensure that critical applications stay available even if one provider experiences an outage.
Customers choose cloud as a secondary/DR environment for their on-premises systems as it reduces the cost and complexity of maintaining a physical DR site. Services like Azure Site Recovery help implementing a disaster environment in the cloud and minimizes the infrastructure cost by spinning up the infrastructure/VMs only during a failover.
Client-driven cloud preferences
Large enterprise customers sometimes have existing relationships with cloud providers and may ask that their deployments align with their chosen provider for consistency, integration with their existing infrastructure, or preferred security protocols.
A niche player in Generative AI-based platform solutions conducted a market survey to prioritize the capabilities for the to-be-built platform. While providing feedback to the survey, customers also emphasized having the platform deployed on their cloud of choice. A cloud-agnostic implementation of the platform thus came in as a top priority requirement for the platform.
Cost-effective LLMs
Many organizations are on (or starting on) an AI/Gen AI adoption journey. They are looking to fine tune the pre-trained Large Language Models with own data to unlock powerful AI-Insights that leads to operational efficiency and potential growth. LLMs require Graphical Processing Units (GPUs) and/or Tensor Processing Units (TPUs) for training and inference demanding powerful hardware like NVIDIA A100.
The operational costs associated with running on public clouds can be significant when it comes to higher end configurations and hardware needs. This has led enterprises to evaluate on-premises/private cloud alternatives, where they can use their own hardware which is essentially a one-time investment/cost as compared to ongoing operational expenses in the cloud. In addition, this approach ensures that sensitive data is not leaving the premises, which is important for highly regulated industries like finance and healthcare.
Navigating the Complexities of Multi-cloud
While multi-cloud has many positive values, there are some difficulties/complexities to consider as well. This section will discuss some of them.
Interoperability
Standardization across platforms stays an ongoing issue, especially in ensuring seamless data transfer, workload balancing, and unified automation between clouds. Kubernetes and containers are helping alleviate this challenge by providing consistent orchestration and portability. Careful adoption of platform-agnostic tools and APIs, along with the standardization of operations using multi-cloud management platforms, are a few effective ways to address interoperability challenges.
Compliance and Security
One of the biggest challenges of multi-cloud is keeping compliance and security consistently across environments. CISOs and security teams are focusing more on unified security policies and tools like Palo Alto’s Prisma Cloud or Microsoft Defender for Cloud to monitor security posture across platforms. These tools not only provide visibility into the security state of multi-cloud workloads but also provide hardening guidelines to effectively improve security posture.
Cost Optimization & FinOps integration
Managing cloud costs can be challenging and adopting a multi-cloud strategy further complicates this. Trying to track, manage, and optimize cloud costs without a centralized tool can quickly turn into a messy and frustrating experience. It’s hard to keep everything organized and under control without clear visibility and automation.
An IT carve-out from a parent company had to move their workloads from both AWS and Azure clouds managed by parent company to own cloud environments. Parent companies received pricing discounts from the very large customers (sometimes referred to as hyperscalers) due to larger footprint on the clouds which the carved-out company didn’t get. This resulted in an increase in the cloud consumption cost in the new environment. Establishing cost visibility followed by cost control became a challenge across multiple Azure subscriptions and AWS accounts. By implementing CoreStack, they gained a single pane of glass to manage CloudOps (cloud operations), FinOps (cloud financial management), and SecOps (security operations) across both AWS and Azure clouds.
Multi-cloud cost management solutions like CoreStack or CloudHealth are becoming integral to balancing performance and cost in a multi-cloud environment.
Performance and Latency Issues
Applications that span multiple clouds may experience performance and latency issues, especially if data must traverse significant distances between regions or providers. Ensuring smooth performance for the workloads deployed across clouds requires careful planning and well-designed architecture.
A financial services client faced latency issues after moving their customer-facing application to Azure while keeping their sensitive databases on-premises to meet strict compliance requirements. This deployment model impacted user experience due to latency issues. To resolve this, they implemented Azure ExpressRoute for a high-speed, private connection, reducing latency significantly. They also used Azure Cache for Redis to store frequently accessed data, cutting down on database queries. For non-critical tasks like analytics, they switched to batch processing during off-peak hours. Additionally, they used Azure Data Factory to sync less-sensitive data to a replica database in Azure, enabling faster local queries.
Monitoring
Monitoring in a multi-cloud setup has challenges like getting a clear view across different platforms, inconsistent metrics, and too many alerts causing confusion. Organizations can use unified monitoring tools for better visibility, set standard metrics for easier comparisons, and implement automated systems to manage alerts and focus on the most important issues.
A customer in the insurance industry has their containerized workloads running on-premises as well as in Azure/AWS. They addressed their observability needs by utilizing Prometheus for metrics collection and Grafana for their visualizations. Prometheus enabled them to efficiently gather real-time metrics across their hybrid/multi-cloud environment, providing detailed insights into container performance, resource utilization, and application health. Grafana complemented this by offering customizable dashboards that unified data from multiple Prometheus instances, delivering a single-pane-of-glass view across their on-premises and cloud workloads.
Multi-cloud Infrastructure Management
Managing infrastructure across multiple clouds requires efficient automation tools. Infrastructure-as-code (IaC) practices allow consistent deployment across clouds. Automation reduces manual intervention, helps in scaling and enforcing governance across multi-cloud environments.
Terraform, an open-source IaC tool developed by HashiCorp, is often preferred over cloud-native/cloud-specific IaC tools like Bicep, Azure Resource Manager templates, CloudFormation because it supports multiple cloud providers (AWS, Azure, Google Cloud), enabling a unified infrastructure management approach.
Key Takeaways
Before considering spreading your computing and data assets across different cloud companies, it is important to be sure and truly understand the risks and complexities along with the value you are likely to achieve. Including:
- A multi-cloud approach can enhance flexibility and resilience but requires a well-thought-out strategy.
- Without a centralized approach to management, organizations risk creating additional complexity and costs.
- Proper monitoring, standardized metrics, and automated alert systems are crucial to address the challenges of multi-cloud environments.
- Regularly assess and refine your multi-cloud strategy to adapt to changing business needs, technological advancements, and evolving regulatory requirements.
However, with this understanding under your belt, you can do great things by using the different cloud platforms in tandem with each other.
