CAST AI, a tool recommended by Delfia, met the company's objectives: cost savings and visibility of Kubernetes, in addition to making applications available
The team of Delfia, curation of digital journeys, implemented the Cast AI resource management solution in the cloud infrastructure of Gazin, a retail giant with more than 365 stores throughout Brazil, which turned 60 in December. With the natural expansion of its business over the years – and with a niche of eleven businesses, including digital banking, insurance, consortium, retail, wholesale, industry, transportation, e-commerce channel, even gas stations and agribusiness – the company needed a partner to support the expansion and reduce expenses, especially in the IT area, since the growth of operations also generates costs.
According to Alex Santos, the company's IT Manager, the tool recommended by Delfia's team of curators to reduce costs in cloud environments keeps Kubernetes optimized. “Gazin is a company that doubles in size every five years. Imagine if we doubled the costs of the environments every five years? I increase profitability, but I can't keep up with expenses in the same way I keep up with profit gains,” says Santos, from Gazin.
The Cast AI tool was used to implement the digital bank, e-commerce (B2C) and wholesale e-commerce – the company's B2B. The Gazin team expected to reduce costs by at least 30% in Kubernetes. Delfia's consulting services in implementing the solution resulted in an average reduction of 40% in the production cluster. Working with spot machines, infrastructures at very low prices and used together with the intelligence of the Cast AI tool, the company maintained the stability of operations efficiently, in addition to saving resources in the environments.
“We had an average of 80% reduction in total project cost considering homologation and test environments”, reveals Edson Macoto Junior, Tech Lead of Gazin's SRE and DevOps teams, explaining that e-commerce, wholesale and banking clusters had a considerable cost. But today, running on spot machines, they have achieved this revenue savings in these non-production environments.
Sustainable expansion
Macoto Junior, from Gazin, says that when the teams started talking about FinOps – an approach to financial management of the use of cloud technology – the discussions about cost reduction were pertinent, but they needed to be sustainable and controlled. “If we evaluate the number of squads we have, implementing new solutions, today we have the practice of working with cloud native applications in a scalable way, and we know that this generates complexity in our environment”.
In addition to resource optimization, Gazin also wanted reliability and scalability. By combining these two fronts, the Cast AI solution met these needs. “Delfia introduced us to the tool, which looks at our workloads, understands the behavior of the applications and provides machines and scaling according to what we need, without having an operational person to command it. It came with a glove for what we needed”, says Gazin’s Tech Lead.
“Delfia made itself available and, through curation, found a solution to Gazin’s need, which was to reduce costs in Kubernetes,” says Plinio Moreira, Delfia’s Observability Manager. In addition, the company’s team of digital journey curation sought to understand the company’s problems. “Visibility of Kubernetes usage was also one of the challenges, which generated excessive expenses. Another point was how the DevSecOps team could act, since there was no solution that could monitor and manage effectively.”
Tool in practice
Delfia’s curation was essential for the Gazin team to make the decision to adopt the solution. “Delfia provides us with a lot of support on the technical side and whenever I need something, everything is ready as well.” Macoto states that a business case plan was put together, the tool was put into operation and the results were collected, mainly from the e-commerce cluster. “It wouldn’t be the same if we just picked the tool and tested it without having done a case study,” says Gazin’s Tech Lead.
“We focus a lot on optimization, which is really the cherry on the cake of the tool, but it also has a whole visualization part: when we instrument the company's infrastructure within the solution, it gives us a complete report of how much is being spent per month, what the potential for savings is and even executes AI-based automations automatically. Full instrumentation is a simple process that takes about seven days, which is the time for the solution's AI algorithms to provide more accurate insights, but on day 1 we already have impressive results of potential savings”, reports Plinio Moreira, from Delfia.
Black Friday 2024 is considered a case study for the Gazin team, in relation to problems they had in the past that were solved with the solution. “And we have larger applications, with a dedicated cluster and another with smaller applications, which run together. There are many applications, with different behaviors, that Cast AI can better understand each one to see if there is a problem. Having stability and in a very healthy way, brings a great return to the company”, says Edson Macoto Junior from Gazin.
Saving time, resources and finances
The operational team also benefited, as there was no longer a daily operational cost to monitor the clusters: the tool already does this. With the use of Cast AI, teams could also focus on meeting business demands instead of making manual adjustments.
“In addition to saving money, which can be reinvested in other areas such as security or observability, the tool also saves a lot of time for the technical team, which, by benefiting from the solution’s AI, no longer needs to perform manual tasks such as rebalancing environments and reallocating instance types (on-demand, spot). The analyst can now focus on the business, developing new products or making security improvements, as the tool already takes care of the optimization automatically,” said Plinio Moreira from Delfia.
Availability was also an important factor for Gazin. The retail company’s team tried to understand which type of machine was best suited to provide each type of service. “New solutions were implemented, or even restructured, changing the application’s behavior. We had to manually try to understand whether machines with more CPUs or more memory, or even more resources, were needed. Cast AI, curated by Delfia, solved this for us, overcoming all these issues, making the applications available. At the same time that we have operational reduction, we have cost reduction and gains in availability”, says Edson Macoto.
Future
In the future, Gazin intends to extend the implementation of Cast AI to other business units. “We will implement it in everything that the tool can support us and we can expand. We started with digital banking and e-commerce and we will implement it in other environments that are starting to make sense, such as Gazin Industry or the insurance company,” reveals Alex Santos. “The level of expansion will be organic, as we are able to implement the tool in new businesses and current ones that we already have,” adds the executive.
Spot machines are part of Gazin's plan for 2025. “For very large applications with many instances, such as e-commerce, for example, which involves many nodes, we work with a percentage of 30%, on top of spot machines, to guarantee availability, because it will be distributed between spot and on-demand machines. And so, we will have a good cost reduction”, highlights Edson Macoto.
Also this year, Gazin’s IT team intends to migrate applications to the on-premise environment. “We will conduct a study with Delfia of other tools, as we want to bring much of what we currently have in the cloud to the two data centers we have in the company. If this solution also makes sense for on-premise, we will start using it. And this will already be a case study for 2025”, concludes Alex Santos, from Gazin.
