Reduction In Cloud Compute Costs

Evidence and data. After interviewees’ organizations began migrating Java services to Platform Prime, they saw a reduction in their cloud costs. Platform Prime optimizes JVM performance and reduces CPU utilization and latency spikes. This means fewer compute instances are needed to process the same workload. The reductions interviewees saw in cloud costs ranged from 7% to 50%.

• The chief information officer at a financial services firm said: “On the financial side, we saw improvements in runtime efficiency and overall optimization. Across some of our Java-heavy workloads, we observed a 7% to 10% reduction in total compute costs. In many cases, that translated to nearly seven figures in annual cloud spend savings. Given that our cloud spend is in the hundreds of millions, even partial migration has already resulted in multimillion-dollar annual reductions. We believe there’s still more savings to be realized as we continue the migration.”

• The VP of digital and data platforms at a financial services firm said, “Through more efficient utilization of our Java applications, we’ve been able to reduce the number of physical and virtual servers in use, resulting in a 15% to 20% reduction in cloud costs.”

• The chief information and technology officer at the public sector company described the reduction in costs in public and private cloud environments: “In public cloud or SaaS-like environments, we see improvements closer to 50%, while air-gapped private clouds are more around 20%, because those environments are less elastic. The reason for such a significant reduction in infrastructure cost is that compute resources are used far more efficiently. Each application runs more efficiently, consumes less, and can handle more transactions on the same hardware. That means that I need fewer resources to process the same workload.”

• The VP of digital and data platforms at a financial services firm experienced long-term savings, even though Platform Prime’s upfront costs may be higher than those of other platforms. “While you’re spending more upfront on Azul Platform Prime compared to other platforms, it delivers savings by reducing the need to expand physical or virtual server footprint.”

Modeling and assumptions. Based on the interviews, Forrester assumes the following about the composite organization:

• The composite organization has 550 Java applications.

• The percentage of Java applications running in the cloud begins at 75% in Year 1 and increases 5 percentage points each year.

• Cloud compute costs per Java application begin at $29,000 in Year 1 and decrease 7.5% each year.

• The total cloud compute cost for Java applications before Platform Prime is just under $12 million per year.

• The migration to Platform Prime and resulting reduction in cloud compute costs is achieved in three phases:

  • In phase 1, the composite migrates 40% of Java applications, resulting in a 27% reduction in cloud costs. In this initial phase, the most resource-intensive applications are migrated first, realizing the larger gains in cost reduction.
  • In phase 2, 30% of Java applications are migrated, and the composite realizes a reduction in cloud costs of 15%. In this phase, the gains begin to taper as the remaining applications are less resource-intensive.
  • In phase 3, the composite migrates the final 30% of Java applications, reducing cloud costs by 7%.

Risks. Forrester recognizes that these results may not be representative of all experiences. The following factors may impact this benefit:

• The number of Java applications running in the cloud.

• The size of the Java applications and the sequencing of the migration of larger Java applications.

Results. To account for these risks, Forrester adjusted this benefit downward by 5%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $4.0 million.

Reduction In Data Center Infrastructure Costs

Evidence and data. Interviewees’ organizations also reported a reduction in on-premises data center costs due to decreasing instance numbers. The reduced need for on-premises servers allowed the organizations to reallocate the hardware to other workloads.

• The VP of digital and data platforms at a financial services firm said: “Although Azul Platform Prime primarily supports our cloud infrastructure, we also have physical servers on-premises that are integrated. As we optimize the cloud environment, those improvements extend to our on-prem architecture as well. This integration reduces the number of physical servers we need to maintain on-prem.”

• The senior manager for performance, resilience, and scalability at the cloud software provider discussed savings benefits: “By using Azul Platform Prime in our private cloud, we’ve been able to reduce costs by about 3x. Its low-latency JVM lets us pack up to three times more customers onto the same machine compared to before, which makes a big difference in efficiency and scalability.”

• The enterprise architect at the technology digital services provider described their capacity planning and their ability to decrease the number of servers required in their upcoming infrastructure refresh: “As we prepare for our infrastructure refresh next fiscal year, we’re planning for future capacity needs, and based on what we’re seeing, we anticipate using roughly 25% to 30% less capacity on these systems overall. There are some nuances here that aren’t directly related to Azul, but I’d estimate that about 15% of the reduction is directly tied to the optimizations Azul provides. That’s a significant impact for us.”

Modeling and assumptions. Based on the interviews, Forrester assumes the following about the composite organization:

• The percentage of Java applications running on-premises begins at 25% in Year 1 and decreases 5 percentage points each year.

• Data center infrastructure compute costs per Java application begin at $21,000 in Year 1 and decrease 7.5% each year.

• The total data center infrastructure compute costs for Java applications before Platform Prime are $2.9 million in the first year, $2.1 million in the second year, and $1.5 million in the third.

• The migration to Platform Prime and resulting reduction in on-premises compute costs are achieved in three phases:

  • In phase 1, 40% of Java applications are migrated, resulting in an 18% reduction in costs. In this first phase, the most resource-intensive applications are migrated first, realizing the larger gains in cost reduction.
  • In phase 2, 30% of Java applications are migrated, resulting in a reduction in costs of 16%. In this phase, the gains begin to taper as the remaining applications are less resource-intensive.
  • In phase 3, the remaining 30% of Java applications are migrated to Platform Prime, resulting in 14% savings on data center infrastructure costs.

Risks. Forrester recognizes that these results may not be representative of all experiences. The following factors may impact this benefit:

• The number of Java applications running on-premises in the data center.

• The size of the Java applications and the sequencing of the migration of larger Java applications.

Results. To account for these risks, Forrester adjusted this benefit downward by 10%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $523,000.

Engineer FTE Reallocation To Higher-Value Work

Evidence and data. Before implementing Platform Prime, customers’ performance engineers spent a significant amount of time handling application performance issues and outages. This included mitigating GC-related disruptions, managing performance volatility, and conducting manual runtime tuning (i.e., adjusting heap sizes and garbage collection types). After implementing Platform Prime, the Java performance was more stable and predictable, resulting in fewer alerts and less time spent troubleshooting and performance turning. This reduction in the amount of time performance engineers spent handling performance issues allowed teams to reallocate engineer FTEs to higher-level work.

• The senior manager of performance, resilience, and scalability at the cloud-based software provider noted a significant decrease in tickets: “Before migrating to Platform Prime, we were handling hundreds of support tickets related to performance issues and tenant outages. Since the migration, that number has dropped significantly. This dramatic reduction has eliminated the need for dedicated engineering teams to troubleshoot JVM-related problems, saving significant engineering hours and improving overall operational efficiency.”

• The chief information officer at a financial services firm described the advantages of a more predictable system: “When you have a more predictable system — fewer alerts and stabilized scaling issues — you reduce the amount of troubleshooting required. Our platform engineers and [site reliability engineering] teams end up spending less time firefighting and more time focusing on productive, value-driven work like automation and observability improvements.”
  The executive went on to describe how performance monitoring was more effective with fewer triggered events: “When you’re looking at performance monitoring — like your application performance management dashboards — you can get a lot of noise. Tons of events pop up, and it takes time for teams to sift through them. If you reduce the number of events triggered by performance issues, you improve your signal-to-noise ratio. That means teams can respond to meaningful events faster and more effectively.”

Modeling and assumptions. Based on the interviews, Forrester assumes the following about the composite organization:

• Thirty-five performance engineer FTEs are dedicated to maintaining JVM performance and mitigating disruptions and outages before Platform Prime, and they spend 80% of their time on these tasks.

• After migrating to Platform Prime, the need to manage JVM performance issues decreases significantly, and performance engineers reduce their effort by 85% in Year 1 and by an additional 5 percentage points each subsequent year.

• These savings are applied to the number of applications migrated over time. The percentage of applications migrated to Platform Prime in Year 1 is 40%. In Year 2, the percentage of migrated applications grows to 70%, and in Year 3, to 100%.

• The fully burdened annual salary of an engineer FTE is $168,000.⁴

• The reduced effort allows engineers to be reallocated to higher-value work.

Risks. Forrester recognizes that these results may not be representative of all experiences. The following factors may impact this benefit:

• The frequency and complexity of performance issues, disruptions, and outages prior to deploying Platform Prime.

• The size of the performance engineer team.

Results. To account for these risks, Forrester adjusted this benefit downward by 10%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $5.7 million.

Unquantified Benefits

Interviewees mentioned the following additional benefits that their organizations experienced but were not able to quantify:

• Impact on customer experience. The implementation of Platform Prime went beyond technical and financial metrics — it also impacted the customer experience. Faster transactions and smoother interactions resulted in improved application performance due to Platform Prime. The chief information officer at a financial services firm stated: “With the improvements we’ve made, transactions post faster, pages load quicker, and the overall experience just feels smoother. You’re getting faster response times, and that makes a noticeable difference in how the customer interacts with the platform.”
  At the gaming technology company, the chief technology officer said the lack of noticeable issues made a difference. They said, “The biggest impact on customer experience is what they don’t notice — no latency, no login issues, and no long waits for results.”
  The VP of digital and data platforms at the financial services firm cited faster foreign exchange transfers: “Platform Prime has led to faster transaction processing, which improves both customer experience and institutional performance. For example, in online banking, a customer conducting a foreign exchange transfer benefits from quicker execution — even if the difference is measured in nanoseconds.”

• Prior JDK license cost savings for customers that switch to both Platform Prime and Platform Core. Some of the customers that switched to Azul’s Platform Prime also made a switch to Azul’s Platform Core product, which is an OpenJDK product line for less mission-critical applications that don’t require the advanced performance optimizations provided by Platform Prime. The combined switch has been instrumental in allowing these customers to migrate all of their Java infrastructure off their prior paid JDK, saving license costs in the process. The chief technology officer at a gaming technology company said, “Azul’s pricing was about 30% less per vCore per year, which was a significant cost improvement over our prior JDK.”
  The enterprise architect at the technology digital services provider said: “The prior JDK provider changed how they were going to be charging for their product. Thinking it was going to be a huge financial impact on our company, we started looking for alternatives, which led us to Azul.”

Flexibility

The value of flexibility is unique to each customer. There are multiple scenarios in which a customer might implement Platform Prime and later realize additional uses and business opportunities, including:

• Cost savings allow redeployment of capital. Cloud cost savings exceeded interviewees’ expectations, which opened up other opportunities. The VP of digital and data platforms at the financial services firm said: “During the pilot phase, we saw a 10% to 12% improvement in cloud and application performance, which formed the baseline of our business case for investing in Platform Prime. However, actual results were up to 20%. This unexpected gain allowed us to redeploy capital beyond what we had planned.”

Flexibility would also be quantified when evaluated as part of a specific project (described in more detail in Total Economic Impact Approach).