How a major telecom provider got a grip on performance, resources and cloud costs

As the cloud environment grew, a new demand arose.

Not: “I don't know what we're paying for the cloud now, but it will be true.”
But: “Is our infrastructure continuously optimally equipped for what our business needs today, and are we in control of the growth of our budget?”

The IT organization noticed that workloads were changing, data was growing exponentially, and teams were setting up resources independently. Performance was stable, but no one could say for sure whether the resource allocation was still in balance with demand, but also with respect to costs.

At Blis Digital, we showed our client two growth forecasts: one if you did nothing and one with the help of the Cloud Optimization process.

That's where our collaboration started.

The challenge: ensuring performance in a dynamic cloud environment

The telecom provider ran business-critical applications in Azure. High availability and performance were absolute preconditions.

What they saw:

• VMs that were spaciously equipped “just to be on the safe side”

• Storage environments that had grown organically

• Log Analytics that were unexpectedly a major cost

• Reserved Instances that weren't being used optimally

• No central overview where resource usage and costs came together

The environment was technically well organized. But the question was:

Can we demonstrate that each workload gets exactly what it needs. No more, no less? And haven't they chosen the safe way to oversize everything here?

And if the latter is the case, how do we optimize it in a controlled manner?

Our Solution: Cloud Resource Optimization as a Managed Service

With this customer, we have our Cloud Resource Optimization Service furnished. The focus was primarily on resource optimization and performance assurance. Cost optimization was the logical but conscious consequence.

Our approach consisted of two technical pillars: insight into actual resource needs and our Cloud Cost Management integration.

Step 1: insight into actual resource needs

We started by integrating IBM Turbonomic into the Azure environment. Turbonomic continuously analyses CPU and memory usage, storage performance and I/O, network load, application demand and performance risks.

Instead of looking at historical averages, Turbonomic works demand-based. Applications “request” resources based on their actual load. The platform then calculates which resource allocation is optimal.

In the first few weeks, we saw:

• Oversizing certain compute resources

• Unattached disks that were easy to clean

• Storage allocations that could be set up more efficiently

• Optimization opportunities in reserved instance coverage and utilization

Step 2: Our Cloud Cost Management Integration

In parallel, we set up our own Cloud Cost Management integration. This is because resource optimization only becomes really powerful when you link it to financial data.

Our Cloud Cost Management integration reads the Azure billing data based on the Focus (FinOps Open Cost & Usage Specification) format. As a result, compute, storage and network data can be analyzed uniformly, costs per resource are traceable and data remains easily multi-cloud expandable. Important: We haven't built our own “mapping” that causes vendor lock-in. We use the open standard as intended.

The choice for the Focus format also makes it possible to expand our service within this customer to other platforms, each with their own billing data structure. With the Focus format, everyone speaks the same language — now and in the future — and this prevents problems with integrations, both organizationally and technically.

Serverless processing and central data hub

Our integration runs on a serverless architecture in which Focus exports are automatically fetched daily, Turbonomic API data is loaded, both datasets are merged and data is indexed in Elastic.

At this telecom provider, for example, we saw that approximately 10% of the monthly bill consisted of Log Analytics, that retention settings and storage strategy were no longer in line with usage and that optimization of logging had a direct effect on resource load and costs. These kinds of insights are only created by technically linking resource and cost data.

To give our customer good insight into the large amounts of data that we collect, we provided a central dashboard as standard. This showed which resources could be optimized, what the potential, realized, planned and rejected savings were. What the effect on performance was, what the financial impact was, how trends developed and where deviations occurred.

With this telecom provider, for example, we saw that:

• Around 10% of the monthly bill consisted of Log Analytics

• Retention settings and storage strategy are no longer in line with use

• Optimization of logging had a direct effect on resource load and costs

You only get these kinds of insights when you technically link resource and cost data.

Our added value: double impact

Our collaboration provided value on two levels.

1 ️ ︎ Technical: better resource balance

• VMs became right-sized without loss of performance

• Reserved instances were strategically redesigned

• Storage has been optimized

• Resource hygiene has been structurally improved

• Engineers received substantiated insights instead of assumptions

Performance remained guaranteed — but with much more insight into how to use the resources, which completely dispelled the idea of “it will be fine”.

2 ️ ︎ Financial: costs in line with business needs

By making optimal use of resources:

• Removed structural overcapacity

• Improved reserved instance utilization

• Did costs become predictable?

• Improved cooperation between IT and Finance

Cost reduction was not the goal, but ultimately the result of efficiently designed and monitored infrastructure.

After two months, we saw hundreds of optimization proposals reported as “potential savings”, many of which were implemented directly by our customer. This collaboration went more and more smoothly, also because implementation was increasingly seen as an integral part of IT Operations. As a result, the realised savings value increased rapidly. Without any negative impact on performance or availability.

Our approach with this telecom provider

We started with a four-week pilot.

Week 1 — Setup & Initial Analysis

• Azure integration

• Setting up data flows

• Identifying quick wins (such as unattached disks)

Week 2—3 — Deepening & Optimization

• Right-sizing proposals

• Reserved instance analysis

• Storage optimization

• First concrete changes

Week 4 — Business case & roadmap

• Management reporting

• Overview of improvements made

• Structural optimization approach

After the pilot, we opted for an annual contract within our managed service, where we, as Blis Digital, offered this service as a SaaS service. This means that we only need access to the customer's systems. We'll do the rest.

Since then, we have been providing:

• Weekly monitoring

• Monthly reports

• Ongoing optimizations

• Forecasting and anomaly detection

• This service's Roadmap

• The Finops Principles Roadmap

What this case shows

In large organizations, the cloud is often technically mature. Nevertheless, growth and change almost automatically result in inefficiency.

Cloud Resource Optimization is therefore not about “cutting corners”.
It's about:

• Continuously ensuring performance

• Aligning resources with actual demand

• Connecting Technology and Finance

• Making growth manageable

At this telecom provider, the combination of real-time workload analysis and our Cloud Cost Management integration has led to a structurally more efficient infrastructure — technically and financially.

Lastly

Cloud environments change daily. Workloads are shifting, pricing models are changing, and data is growing. Without active optimization, inefficiency creeps in.

With our Cloud Resource Optimization Service we help organizations to keep their infrastructure continuously in balance. So that performance remains guaranteed and costs are automatically in line with business needs. And all of this is completely transparent.

Curious about what optimization potential is hidden in your cloud environment?
We would like to start with a 4-week pilot.

‍