Maximizing data center performance by getting cloud migration right

This article is part of a VB special issue. Read the full series here: Thefuture of the data center: Handling greater and greater demands.

CIOs and IT infrastructure leaders face a challenging series of decisions in choosing which workloads migrate from data centers to the cloud and which stay on-premise. Combining data-driven insights from generative AI and machine learning (ML) with contextual intelligence from experts can help. 

Developing cloud migration roadmaps that capture what gen AI recommends, factored by contextual intelligence from experienced IT experts, delivers the most reliable results. How effective gen AI and ML are at enhancing data center performance depends on the quality of infrastructure supporting it. CIOs tell VentureBeat that they’re under increased pressure to get more data center consolidation done with less budget and often smaller teams while increasing their performance. 

Nvidia seizes data center opportunity

Nvidia jumped on the data center opportunity early because CIOs and their teams are short-handed and need technologies that can securely scale with fluctuating workload levels while delivering performance gains and reducing costs. The company’s data center strategy concentrates on delivering the performance, sustainability and operating cost reductions. 

Data centers are now NVIDIA’s largest business, at 56% of FY23 revenue. For Fiscal 2024 Q1, Nvidia reported quarterly revenue of $7.19 billion, up 19% from the previous quarter. Nvidia’s record first-quarter data center revenue of $4.28 billion, up 14% from a year ago and 18% from the previous quarter, reflects the strong demand from enterprises for using AI and ML-based technologies to improve data center performance. 

Getting more value out of data centers the goal

Finding new ways to reduce data center costs while improving performance is a high priority for tech leaders. CIOs say their boards of directors are holding back on capital expense (CAPEX) spending for new data center improvements, shifting to a more operating expense (OPEX) based strategy, which is common with a cloud-centric infrastructure.

CIOs in financial services also say that the workloads with the most sensitive financial data, including transactions with regional federal banks, must stay on-premise. It is often less expensive than moving the workloads to the cloud. According to an Nvidia report, 44% of financial services firms rely on a hybrid infrastructure for their AI workloads and projects. 

One strategy that’s working is identifying how improved data center operations contribute to greater sustainability. With the CEO and other senior management team members seeing their total compensation programs indexed to environmental, social, and corporate governance (ESG) plans, CIOs tell VentureBeat that tying data center modernization to corporate-wide initiatives helps.

Pursuing sustainability initiatives

Gaining a budget to pursue sustainability initiatives quickly becomes a core part of every CIO’s cost-reduction strategy for data centers as they respond to rising energy costs, supply constraints and uncertain economic conditions. Reducing excess power, investing in clean energy and delaying replacement cycles are crucial for attaining this goal. 

Cloud or colocation services help CIOs consolidate data centers and close unnecessary facilities. Public cloud and colocation providers prioritize sustainable computing and clean energy to attract new data center businesses.

Gartner recently found that enterprises could achieve up to 60% in cost savings by using sustainability-based initiatives to extend server life spans from three to five years. By combining AI, gen AI and ML techniques to analyze real-time server data, enterprises achieve higher server utilization, storage capacity and greater visibility and control over operating costs.

Getting cloud migration right is difficult

Getting the business case and technical roadmaps right for cloud migration is complex. CIOs tell VentureBeat that it’s often an iterative process and advise thinking of it in the context of a digital transformation of a business beyond a cost-cutting strategy.

By 2025, Gartner predicts that more than 95% of new digital workloads will be deployed on cloud-native platforms, up from 30% in 2021.

“There is no business strategy without a cloud strategy,” said Milind Govekar, Gartner distinguished VP analyst. “New workloads deployed in a cloud-native environment will be pervasive, not just popular, and anything non-cloud will be considered legacy.”

Gartner defines cloud migration as “the process of planning and executing the movement of applications or workloads from on-premises infrastructure to external cloud services, or between different external cloud services. At a minimum, applications are rehosted (moved largely as-is to public cloud infrastructure). Still, they are ideally modernized through refactoring or rewriting, or potentially replaced with software as a service (SaaS).”

AWS Migration Hub, Google Cloud Migration and Microsoft Azure Migrate are designed to help IT teams with the migration process and ongoing management of cloud workloads. Azure Migrate provides a business case builder app with step-by-step instructions that display comparisons between on-premises and Azure total cost of ownership, year-on-year cash flow analysis and resource utilization-based insights to identify servers and workloads well suited for the cloud. 

Cloud migration roadmaps need a solid purpose to succeed

C-level executives who have led successful cloud migration strategies tell VentureBeat it’s best to take a long-term perspective and assume that it will take up to two times as long as initial estimates. 

The reason: Strong resistance to change. CIOs tell VentureBeat that it’s imperative to be completely transparent on the goals behind each specific roadmap, which new business initiatives it’s supporting and if cost reduction is also a goal.

Cloud migration roadmaps need to address five key areas to be effective in offloading workloads that weigh down data center performance. These include assessing workflows and their cost, performance and security implications before migration. Secondly, IT, operations and the office of the CIO need to define which cloud platform provider(s) makes the most sense for given workloads, migration strategies (from rehosting to replacing), implementation plans to minimize disruption and how IT can optimize workloads through continuous monitoring and adjustment of cloud resources.

There are five most common migration strategies, which can make or break a cloud migration. The first, as defined by Gartner, is Rehost or “lift and shift,” which involves moving an application from one platform or IT environment to another.

Replatform, or “lift and reshape,” involves revising an application’s architecture while preserving its core functionality. Rearchitect refers to reengineering or refactoring an app’s architecture, and Rebuild refers to rewriting or redesigning an application from scratch. Finally, Replace refers to repurchasing a new solution or “dropping” the old one and “shopping” for a new one.   

Cloud migration roadmaps have the highest probability of success when the anticipated changes to each application are understood, and the considerations and impact of each cloud migration strategy are defined. Source: Gartner.

Measuring results

The metrics and KPIs each CIO and their team choose to monitor as part of a cloud migration strategy will be determined by the business model they’re trying to make more efficient, legacy system integration workloads, constraints in moving specific systems, budgets and available teams to work on the project.

Across the CIOs VentureBeat has spoken with, a core set of metrics are standard across most cloud migration. The most important factor is improving the user experience by increasing system responsiveness, reliability and scale to support unpredictable resource loads.

The second is application performance; the third is keeping performance baseline comparisons accurate with real-time monitoring. Additional metrics include service-to-service latency, server performance (which can identify hidden effects not previously identified, and error rates and response time valuations.

Reparation of cloud workloads happens when cloud migration strategies don’t deliver the promised performance or cost gains. Having a repatriation plan in place is now common in the highest-risk industries, notably banking, insurance and financial services that need to have a fallback plan immediately if cloud migration runs into unexpected delays or problems.