Groundcover acquisition automates Kubernetes provisioning

An emerging observability vendor's first acquisition will combine its observability data with a specialized Kubernetes resource-optimization algorithm, setting it up as a 'platform of action' for users contending with a fast-paced, AI-driven world.

The acquisition of Wand, based in Tel Aviv, Israel for an undisclosed amount by Groundcover reflects a broader trend toward autonomous infrastructure among vendors in the observability market. Observability vendors such as Dynatrace and Datadog are also using context in their observability data to automate infrastructure management, rather than simply reporting on it.

While those companies mainly use that context to diagnose and remediate issues with application code, Groundcover plans to use it to autonomously optimize infrastructure, according to CEO Shahar Azulay.

"There's a lot of knowledge in what the Wand team has built on how to optimize Kubernetes and cloud workloads continuously with the use of deep algorithms that can dynamically scale the environment – which is where we believe the world is going, rather than static, engineering-made decisions about [how] the infrastructure should serve the application layer," Azulay said.

Like applications, infrastructure is also becoming autonomous, he said.

"It's going to be controlled by agents, [and] we'll adapt it dynamically to the needs of the application. Observability platforms like us have the entire context," Azulay said. "We see what's going on from the application layer… so we can feed algorithms [like Wand's] with the context of what's actually going [on] in production."

The move beyond observability into infrastructure automation is driven by AI workloads and the maturity of cloud-native technology adoption, said Stephen Elliot, an IDC analyst.

Stephen Elliot

"Customers recognize that they need a lot of data to understand the performance of their services. And [increasingly] with the pressure to drive AI ROI, they've got to get out a contextual understanding of what's happening with agents … across these infrastructures," Elliot said. "It's not just about a handful of agents anymore. It's actually about hundreds or thousands of agents that are going to be working together as a system."

Vendors and customers are realizing these systems are no longer optional, and observability tools have evolved from reactive to proactive to autonomous, Elliot said.

"It's no longer good enough just to collect a massive amount of information and parse through it," Elliot said. "[Vendors] are now saying, 'We're going to collect a lot of information for you, but really, this is what's going on and this is what you should do to solve it. And if you would like us to solve it, we can solve it for you and drive a more autonomous system.'"

Companies are increasingly looking to automation to help with infrastructure cost and reliability, Azulay said.

"There's such a big pain [when it comes to the] cost and reliability of the infrastructure. It's no longer a layer you can discard and say it shouldn't work. It's part of how the application scales and operates," Azulay said. "Modern companies think of the infrastructure and application together."

AI magnifies these cost concerns for development and platform teams, as well as C-level executives.

"For AI return on investment, there is more pressure for everyone, including developers and AI engineers, to understand the cost of their decision making," Elliot said.  

Context for Kubernetes-specific optimization

The Wand acquisition will provide Groundcover users with that data, context and way to act specifically on Kubernetes infrastructure, according to Elliot.

"[Wand's] got a level of depth, a level of focus, [and] very specific information on Kubernetes deployments that [can] drive a level of recommendations, insight, and potentially autonomy that maybe customers can't get from other vendors," Elliot said. 

Wand addresses a common problem in Kubernetes implementations. Teams often isolate workloads or provision excess capacity because they don't always know how applications that share infrastructure will affect one another, said Torsten Volk, an analyst at Omdia, a division of Informa TechTarget.

For all you know, that same Kubernetes cluster has 25 other nodes that do nothing. [Wand] looks at the workloads over time, how much they absorb and how that impacts their performance.

"If you run multiple applications on the same node, without analysis from something like Wand, you have no idea when something goes slow, if it was one application or the other application [that caused it]," Volk said. "For all you know, that same Kubernetes cluster has 25 other nodes that do nothing. [Wand] looks at the workloads over time, how much they absorb and how that impacts their performance."

Torsten Volk

Wand juggles workloads based on their impact on the overall infrastructure and on one another, Volk said. "If one of them tends to be a noisy neighbor, it gets separated, or it gets scaled up so that it has more memory and doesn't disturb the others."

That ability to account for the behavior of multiple workloads is what Groundcover sees as a step beyond simpler approaches to Kubernetes resource optimization, Azulay said.

“[Wand] is taking it a step further in the way they reallocate applications across the infrastructure, regrouping applications across different servers to create a much more dynamic span of how they can actually optimize the infrastructure,” Azulay said.

The data that Groundcover collects as a bring-your-own-cloud provider, combined with its eBPF-based sensor, gives it a clear picture of the customer environment to feed Wand's algorithm. Its positioning in the stack means Groundcover can provide Wand with a detailed view of an application and the layers of Kubernetes and cloud infrastructure on which it runs simultaneously, Azulay said.

"I can see what APIs are being called and who's using what without being part of the code," Azulay said. "I'm running on the infrastructure, so I can see infrastructure metrics like the CPU utilization to a very high granularity, and memory utilization, and still see the application stack at the same time with the same sensor."

Ben Lutkevich is an award-winning writer for TechTarget.com.