How to Optimise Multi-Cloud Workload Performance

Managing workloads across multiple cloud platforms can help you improve performance, reduce costs, and strengthen security. However, it comes with challenges like rising expenses, complexity, and compliance risks. Here's how you can tackle these issues:

• Discover and Map Workloads: Use monitoring tools and automation to classify and tag applications, ensuring better resource allocation and cost tracking.
• Monitor Performance: Track key metrics like latency, CPU usage, and error rates to identify and resolve issues quickly. Set real-time alerts for proactive management.
• Automate Scaling: Implement autoscaling and spot instances to dynamically adjust resources based on demand, saving up to 90% on costs.
• Distribute Workloads: Strategically place applications across clouds to reduce latency, improve resilience, and meet local regulations.
• Optimise Networks: Use direct connections, SD-WAN, and traffic prioritisation for faster and more secure data transfer.
• Leverage Tools: Platforms like Datadog, Terraform, and CloudZero simplify monitoring, automation, and cost management.

The Multi-Cloud Expedition Episode 11: Optimizing Multi-Cloud Workload Placement

Workload Discovery and Classification

Before you can improve performance across multiple cloud platforms, you need to understand what you're dealing with. Accurate workload discovery and classification are essential for aligning infrastructure and cloud services with business needs [6].

The Flexera 2025 State of the Cloud Report reveals that 89% of enterprises adopt multi-cloud strategies to reduce risks and improve flexibility [4]. However, a study by Virtana found that 82% of businesses with public cloud workloads face unnecessary expenses due to poor resource visibility [10]. This underscores why proper workload discovery is so critical.

Multi-cloud management is more than a buzzword - it's a strategic imperative. - Edward Ionel, Head of Growth, Mirantis [7]

To understand workloads, you need to examine applications, their dependencies, and latency requirements. This analysis helps determine the best placement across different cloud deployment models [12]. It also ensures cost accountability across teams and reveals usage patterns that directly affect your bottom line [13]. With this groundwork, you can confidently map and tag workloads in multi-cloud environments.

How to Discover and Map Workloads

Mapping workloads involves systematically identifying and cataloguing every application, service, and data flow within your multi-cloud setup. The aim is to gain complete visibility, allowing for smarter resource allocation and performance improvements.

Start by using a unified monitoring and management solution. This centralised approach ensures visibility and control during workload and data migrations, avoiding fragmented management across various cloud providers [5].

Automation frameworks are also key. These tools can manage workloads by placing them in the right cloud environment at the right time, based on business policies, cost limits, and performance needs [5]. Manual tracking simply isn’t practical in dynamic environments - automation ensures continuous discovery and mapping as systems evolve.

For example, Workload Discovery on AWS illustrates how discovery tools work. It offers visualisation of AWS workloads, maintains an inventory of resources, maps their relationships, and presents everything through a web-based interface [3]. Similar tools are available for other cloud providers, ensuring unified visibility across platforms.

Independent monitoring tools are particularly valuable in multi-cloud setups. They provide consistent visibility, no matter where your workloads are hosted [6]. The goal of discovery is to fully understand workload demands, ensuring cloud resources are reliable and efficiently utilised [12].

Workload Tagging and Labelling

Once workloads are mapped, tagging becomes essential for managing resources and tracking costs effectively. Tagging standardises inventory management, supports cost allocation, enforces policies, and enables automation within cloud environments [10].

To create a robust tagging system, use a standardised, case-sensitive format that works across all cloud providers [8]. Each platform has its own tagging rules, so finding common ground is crucial. For instance:

When starting, apply as many tags as your platform supports - AWS, Azure, and GCP all allow at least 50 tags per resource [8]. It’s easier to remove unnecessary tags later than to retrofit a tagging strategy across a large number of resources.

A strong tagging strategy should include these five categories [9]:

• Functional tags: Labels like app:catalogsearch1, tier:web, env:prod, or region:uksouth make it easy to identify workload types and environments.
• Classification tags: Support governance and security with tags such as criticality:mission-critical, confidentiality:private, or sla:24hours. These help prioritise resources during incidents.
• Accounting tags: Enable cost tracking with labels like department:finance, costcentre:55332, or budget:£200,000. These are essential for chargeback and showback processes.
• Purpose tags: Align resources with business goals using tags like businessprocess:support or revenueimpact:high. These help justify cloud investments and prioritise spending.
• Ownership tags: Ensure accountability with labels such as businessunit:finance, opsteam:central-it, or opsteam:cloud-operations. This makes it clear who is responsible for resource performance and costs.

One pharmaceutical company successfully implemented a detailed tagging strategy tailored to their R&D workflows. They used granular tags like ResearchArea, ProjectID, and WorkloadType, combined with standard naming conventions and integration with R&D tools. This approach led to precise cost tracking, better collaboration, and more informed decision-making [11].

To enforce tagging, use Infrastructure as Code (IaC) templates, CI pipelines, or service control policies [10]. Regularly monitor tag coverage and assign resource owners to maintain accountability.

Finally, ensure consistency in naming conventions, labels, or tags to categorise workloads by team, application, environment, or cost centre [13]. This structured approach not only clarifies your resource inventory but also highlights how each resource contributes to business goals, helping you identify areas for improvement.

Performance Monitoring and Metrics

Building on workload mapping and tagging, effective monitoring brings together performance, cost, and security metrics into one cohesive system. Without a clear, unified view across all cloud environments, decisions about optimisation can end up being based on incomplete or misleading data. This comprehensive perspective forms the foundation for tracking the critical metrics outlined below.

Cloud monitoring serves as the glue that connects resource tracking, cost management, performance evaluation, and security improvements [17]. In multi-cloud environments, ensuring standardisation of data across providers is crucial [14]. Centralised observability platforms simplify this process by consolidating data from applications, networks, and cloud technologies into a single, easy-to-access interface [16]. This unified monitoring approach helps avoid fragmented oversight, which can lead to delayed problem resolution and a diminished customer experience [20].

Key Metrics to Track

To monitor effectively in a multi-cloud setup, it’s essential to focus on the metrics that matter most. The Golden Signals framework offers a great starting point, covering four key areas: Latency, Traffic, Errors, and Saturation [14]. Beyond these, other important metrics include CPU and memory usage, request rates, response times, error rates, network latency, packet loss, and system availability [15]. By prioritising metrics that directly affect business outcomes, teams can avoid being overwhelmed by unnecessary data [17].

For multi-cloud systems, tracking metrics at both regional and provider-specific levels can help identify latency issues [15]. Establishing baseline patterns for resource usage is equally important, as deviations from these norms can signal security breaches or performance bottlenecks [14]. Collaboration between teams is crucial here: development teams bring expertise in application behaviour, while operations teams understand infrastructure trends [17].

Real-Time Monitoring and Alerts

Once the key metrics are in place, real-time alerts become the next line of defence. These alerts allow for proactive management by flagging potential issues before they impact users [22]. For UK businesses, configuring alerts to align with local requirements is straightforward - most monitoring tools allow for custom date and time settings, ensuring alerts are triggered accurately within the UK’s time zone [19]. For example, critical alerts might be active 24/7, while less urgent notifications could be limited to standard business hours (09:00–17:30 GMT).

Alerts should be configured with clear thresholds (e.g., CPU usage exceeding 80% triggers a warning, while 95% signals the need for immediate action), multiple notification channels, and tailored schedules [18]. Automation plays a key role in reducing manual errors and enforcing security protocols [21]. Ideally, alerts should also trigger automated responses, such as scaling resources, redirecting traffic, or initiating failover procedures. Consolidating logs and alerts into a single dashboard can greatly simplify troubleshooting. A unified dashboard provides a centralised view of all cloud resources [22], while packet-level insights enable detailed network troubleshooting and early problem detection [20].

For UK businesses operating across multiple time zones, it’s helpful to adjust alerts based on regional peak hours. For instance, a retail company serving customers in both the UK and Asia might set different alert thresholds for its London infrastructure during Asian business hours (01:00–09:00 GMT).

Shifting from isolated monitoring tools to a connected, cross-environment strategy is essential [20]. This integrated approach allows organisations to quickly identify whether an issue is limited to one cloud provider or spans the entire multi-cloud setup. With this clarity, businesses can make informed optimisation decisions and ensure a seamless user experience.

Automated Resource Allocation and Scaling

Once performance patterns are identified through effective monitoring, the logical next step is automation. This eliminates the need for manual adjustments and ensures systems respond dynamically to workload demands. Automation becomes crucial when workloads are fully mapped and monitored, especially in multi-cloud setups. For UK businesses, where traffic often follows predictable patterns tied to standard working hours, this can make a significant difference.

To avoid costly mistakes, it's essential to assess business needs and ensure automation tools align with scalability demands [1]. This step helps define what’s required before choosing tools that work across various cloud platforms.

Autoscaling and Spot Instances

Autoscaling is a game-changer for managing resources effectively. It adjusts resource use based on demand, ensuring cost efficiency while maintaining reliability and scalability [24]. When paired with spot instances, the savings can be substantial - up to 90% compared to on-demand pricing [23].

Take the example of a C5.xlarge spot instance: it costs around £0.0388 per hour, compared to approximately £0.17 per hour for on-demand pricing [23]. However, since spot instances rely on surplus capacity, they come with a catch - they can be interrupted with little notice (e.g., 2 minutes on AWS). With an average interruption rate of 5%, these instances are best suited for workloads that can tolerate interruptions [26].

For UK businesses, leveraging spot instances during off-peak hours (22:00–06:00 GMT) can be particularly cost-effective. Tasks like batch processing, development and testing, CI/CD pipelines, and big data processing can all handle occasional interruptions without issue.

To maximise the benefits of spot instances, applications should be stateless, and tools like AWS Autoscaling Groups should be used to automatically replace interrupted instances. Configuring multiple instance types within each autoscaling group further ensures availability. AWS also offers tools like EC2 Fleet and Spot Fleet, which deploy a mix of instance types across different Availability Zones. Capacity Rebalancing, another useful feature, replaces instances at risk of interruption proactively [23][25].

Automation Tools and Strategies

Beyond autoscaling, integrated automation tools simplify managing resources across multiple environments. Multi-Cloud Management Platforms (CMPs) provide a unified interface to coordinate resources across different cloud providers. With multi-cloud adoption growing, choosing the right tools is critical to maintaining oversight [28].

Infrastructure-as-code (IaC) tools, like Terraform, enable consistent infrastructure changes across providers with version control and automated testing. For instance, a global e-commerce company used Terraform to standardise deployments across AWS and Microsoft Azure, reducing the time needed to launch regional storefronts from weeks to days [28]. This consistency ensures rapid scaling when required.

Configuration management tools complement IaC by automating tasks like application deployment and maintenance. An example? A BFSI firm used Ansible to automate patch management across its hybrid cloud environment, cutting security update times by 70% and improving compliance [28]. Ansible’s agentless nature makes it ideal for multi-cloud setups, where installing agents across platforms can be cumbersome.

For UK businesses, aligning automation with local requirements is important. For example, setting CloudWatch to 'eu-west-2', using £ for costs, and adopting DD/MM/YYYY dates with GMT timestamps ensures local relevance [27]. Scaling policies should also account for UK working hours and public holidays [27]. Analysing historical traffic data can uncover patterns that support predictive scaling strategies.

In October 2023, AWS Blu Insights enhanced its scaling capabilities by integrating ECS, Application Auto Scaling, and CloudWatch. This resulted in faster, more accurate step scaling [27].

Scaling container deployments, it has to start with an application first mindset. - Nathan Peck, Senior Developer Advocate at AWS [27]

AI and machine learning are further transforming multi-cloud automation. AI Copilots now assist DevOps teams with tasks like provisioning infrastructure, enforcing policies, and detecting anomalies. By next year, 75% of organisations are expected to prioritise technology partners that deliver consistent deployment experiences across cloud, edge, and dedicated environments [28].

For those seeking expert advice, Hokstad Consulting offers specialised services in DevOps transformation and cloud cost engineering. They help businesses lower cloud expenses through strategic automation and efficient resource allocation across multi-cloud platforms. These automation strategies, combined with earlier steps like workload monitoring, create a well-rounded optimisation approach.

Workload Distribution Strategies

Once automated resource allocation is in place, the next step is to strategically distribute workloads across multiple clouds. This approach not only meets the UK's data protection regulations but also ensures low latency for users.

Tailored workload distribution has a direct impact on performance, resilience, and costs. For UK businesses, data protection and maintaining low latency for domestic users are key priorities. With around 46% of UK organisations expected to use multiple public clouds within the next three years [2], a well-thought-out workload distribution strategy is becoming increasingly critical.

Multi-cloud strategies allow our clients to select the optimal services from each provider rather than accepting compromise solutions from a single vendor. Businesses have reduced infrastructure costs by 25-40% whilst improving performance and reliability. – Ciaran Connolly, Director at ProfileTree [31]

Strategically distributing workloads can cut cloud expenses by 15-30%, thanks to the pricing advantages offered by different providers [31]. However, achieving these benefits requires careful planning and implementation.

Comparing Distribution Strategies

Each workload distribution strategy serves different business needs. Understanding the trade-offs can help you choose the approach that aligns best with your goals.

Active-active distribution is ideal for applications requiring high availability. Financial services, for example, often use this setup to ensure trading platforms stay operational even during outages. By routing traffic through multiple regions, including UK-based data centres, these businesses can meet regulatory standards without sacrificing performance.

Active-passive setups are better suited for organisations focused on cost control rather than maximum uptime. For instance, a retail company might operate primarily from AWS's London (eu-west-2) region, with a standby environment in Azure UK South. This approach keeps costs manageable while still ensuring business continuity in case of major disruptions.

Geo-distribution combines performance optimisation with compliance. Companies serving European customers might host workloads in London for local users while using Frankfurt or Dublin for broader European coverage. This approach not only ensures GDPR compliance but also minimises latency for users in different regions.

Real-world examples showcase the effectiveness of these strategies. TIM Brasil, for instance, migrated 8,000 workloads and 16 petabytes of storage to a multi-cloud setup, cutting customer service inquiry handling times by 50% on average [29]. Similarly, Liantis in Belgium restructured its multi-cloud architecture to achieve quicker response times, stronger security, and cost savings [29].

We typically recommend starting with a hybrid approach - keeping core systems on your primary provider whilst experimenting with specific workloads on secondary providers. This allows businesses to gain experience and confidence before making major architectural changes. The key is starting with low-risk applications that can provide immediate value. – Ciaran Connolly [31]

The choice of strategy has a direct impact on network performance, which is the next area to address.

Network Optimisation for Multi-Cloud

After selecting a distribution strategy, optimising network performance becomes essential to fully realise its benefits. For workloads spread across multiple cloud providers, network performance can make or break the setup. With IT downtime costing an average of £4,500 per minute - over £270,000 per hour [30] - the stakes are high.

Effective multi-cloud networking involves strategic planning for connectivity, routing, and security. While 85% of organisations use two or more IaaS providers [32], many encounter challenges with network complexity and maintaining performance across platforms.

Direct connectivity is the backbone of optimised multi-cloud networking. Services like AWS Direct Connect, Azure ExpressRoute, and Google Cloud Interconnect offer dedicated connections that bypass the public internet. For UK businesses, establishing direct links to London-based data centres can reduce latency and enhance security for sensitive workloads.

Technologies such as software-defined networking (SDN) and SD-WAN provide additional capabilities. These tools enable centralised control and dynamic routing. For example, a manufacturing company might use SD-WAN to connect its Birmingham headquarters to AWS's London region and Azure UK South, ensuring consistent performance across both platforms.

Traffic prioritisation is another critical factor. Implementing Quality of Service (QoS) policies ensures that essential applications receive adequate bandwidth, while less critical traffic uses remaining capacity. Aligning QoS settings with local working hours (09:00–17:00 GMT) can help manage peak traffic periods effectively.

Proximity placement also plays a big role in performance. Keeping related workloads geographically close reduces data transfer costs and latency. Features like Azure's proximity placement groups and similar tools from other providers can help maintain optimal performance.

When it comes to security, integration across multiple cloud networks requires a comprehensive approach. Virtual firewalls, load balancers, and intrusion detection systems must function seamlessly across all providers. For organisations handling sensitive data, end-to-end encryption and adherence to GDPR guidelines are non-negotiable.

Hokstad Consulting specialises in multi-cloud network optimisation and strategic cloud migrations. Their expertise helps businesses achieve cost savings while maintaining high performance and compliance.

Continuous monitoring is essential to keep multi-cloud networks running smoothly. Tracking metrics like latency, throughput, and packet loss across providers establishes performance baselines and helps quickly identify and resolve issues. Regular fine-tuning ensures the network remains efficient and reliable.

Tools and Best Practices

Once network optimisation and workload distribution are in place, the right tools and methods help refine these efforts. They streamline monitoring, automation, and governance, making a real difference between a multi-cloud setup that thrives and one that drains resources unnecessarily. For instance, research shows that inefficiencies can cost companies up to £20 billion - or 33% - of their cloud budgets [33]. In the UK, where cost control and regulatory compliance are critical, selecting the right tools is essential for balancing performance with financial efficiency.

Key Tools for Optimisation

Successful multi-cloud setups start with effective monitoring tools that offer a unified view of your infrastructure. With nearly 89% of enterprises adopting multi-cloud strategies [4], platforms that consolidate and analyse data across various clouds are indispensable for real-time insights [33].

Datadog is a standout choice, offering extensive integrations and pricing that starts at around £15+ per host per month [34][35].

We have deployed Datadog for all our cloud deployments in AWS. A large number of integrations allow us to literally monitor everything. [34]

For automated discovery, LogicMonitor provides preconfigured templates at approximately £22 per resource/month [34].

Instead of telling your monitoring tool what to monitor, LogicMonitor discovers a lot of metrics and data points for you, mostly out of the box, and away you go. [34]

Dynatrace, priced at around £0.04 per hour for any host size, excels at root cause analysis with AI-driven insights [34].

The Problems App is my personal favourite feature within Dynatrace. It provides a quick summary of the issue, the time it occurred, and a link to the impacted resources so you can dig deeper. [34]

Cost management tools are equally important, especially since nearly 70% of organisations face cloud misconfiguration issues [33]. CloudZero, for instance, has helped companies like Drift cut annual cloud expenses by £1.9 million and enabled Validity to reduce time spent on cloud cost management by 90% [33].

Multi-cloud management platforms should offer features like cost anomaly detection, budgeting, forecasting, and consolidated visibility across providers [4]. These capabilities are becoming increasingly critical, as Gartner predicts global cloud revenue to hit £580 billion by 2025 [4].

Automation tools also play a key role. Platforms like Terraform and AWS CloudFormation enable consistent deployments across clouds, while container orchestration tools like Kubernetes and CI/CD pipelines ensure seamless application delivery.

For UK businesses, firms like Hokstad Consulting provide expertise in DevOps transformation and cloud cost engineering. Their services focus on cutting cloud costs by 30–50% and improving deployment cycles with automated CI/CD pipelines and comprehensive monitoring solutions.

Here’s a quick comparison of tools by category and use case:

While tools are essential, they’re just part of the equation. Consistent operational practices are what keep multi-cloud environments running smoothly over time.

Best Practices for Continuous Improvement

Setting up the right tools is just the start. To maintain and enhance multi-cloud performance, organisations need structured practices that evolve with their environment. As 98% of IT decision-makers plan to use multi-cloud by 2024 [36], ongoing optimisation is key to staying competitive.

Strong governance and alignment are crucial. Establishing a Cloud Centre of Excellence (CCoE) that integrates Governance, Risk and Compliance (GRC), security, cloud platform engineering, and product management ensures consistent decision-making across platforms like AWS, Azure, and GCP [36].

Developing a FinOps discipline is equally important for cost transparency and informed decision-making at the executive level [36]. Regular performance reviews and optimisation cycles should be part of the process. Quarterly assessments of workload placement, cost efficiency, and performance metrics can uncover new opportunities for improvement [1].

Automation is another cornerstone of effective multi-cloud management. Automating repetitive tasks, using Infrastructure as Code (IaC) blueprints [38], and creating integrated virtual environments can reduce errors and improve deployment reliability [39].

Security and compliance must also be prioritised. Automated policy enforcement ensures that speed and control aren’t compromised. Leading organisations benchmark regularly and apply consistent policies across clouds [36]. Data consistency can be achieved using middleware or integration tools to synchronise information across environments [37].

Workload optimisation involves playing to the strengths of each cloud provider. Instead of treating all clouds the same, organisations should evaluate providers based on factors like data sensitivity, workload volume, and specific business needs [37].

For those seeking expert guidance, Hokstad Consulting offers tailored solutions for cloud migration and optimisation. Their No Savings, No Fee approach ensures clients see tangible results.

Finally, measurement and feedback loops are essential for long-term success. A single control plane can help organisations monitor metrics like cost per workload, performance improvements, and security incident reductions [38]. Regular benchmarking against industry standards can highlight areas needing attention, keeping optimisation efforts on track.

Conclusion

Optimising multi-cloud workload performance requires a calculated approach that balances the complexity of managing multiple platforms with the need for efficiency. With the multi-cloud management market projected to hit £15.2 billion by 2027 [40], organisations across the UK must adopt structured strategies to fully leverage their distributed infrastructures.

To start, focus on comprehensive workload discovery and unified monitoring. From there, implement automation for resource allocation and intelligent workload distribution. Research indicates that AI-driven solutions can reduce cloud infrastructure costs by up to 30% [40], making automation not just beneficial but essential. This technical groundwork supports informed, localised decision-making.

For UK businesses, aligning strategies with local requirements like GDPR compliance and data sovereignty is crucial. A hybrid model can serve as a stepping stone, allowing organisations to expand gradually while minimising complexity and building internal expertise.

Cloud orchestration and automation are essential for managing cloud operations effectively. They work hand in hand to ensure that cloud-based tools, applications, and services are set up and maintained correctly, keeping cloud environments running smoothly and working together as a whole. These methods also help businesses improve efficiency, minimise mistakes, and easily scale their operations. [41]

While 82% of business leaders identify AI and cloud computing as key to agility, 95% remain concerned about multi-cloud security risks [40]. This underscores the importance of continuous improvement and robust governance frameworks.

Success in this area depends on structured planning, widespread automation, and ongoing optimisation. With Gartner predicting that 90% of enterprises will adopt multi-cloud environments by 2027 [42], those who master these fundamentals now will gain a significant edge in an increasingly cloud-reliant economy.

Investments in multi-cloud optimisation often lead to cost reductions of 15–30%, alongside improved performance and resilience [31]. For UK organisations, combining strategic planning with the right tools and expert support - such as from Hokstad Consulting - can turn multi-cloud challenges into opportunities. By following the strategies outlined here, businesses can not only mitigate risks but also achieve meaningful cost savings and performance improvements.

FAQs