Reducing Compute Costs by 30% While Managing High-Volume Monitoring Data with IBM Instana Observability

Instana helps companies monitor applications and offers anomaly detection, user experience monitoring, and automatic application performance tracking. The observability tool helps customers make sense of their data and respond quickly to potential issues through timely alerts. “We gather information about everything our customers have deployed in near real time, stream it to a backend system that makes sense of all the relationships and data, and give them a full visualization to uncover any problems the moment they happen or, in some cases, ahead of time,” says Chris Bailey, chief technology officer for Instana.

The key areas the company wanted to improve were scalability and cost-efficiency. These are critical aspects of all software-as-a-service products, which must scale seamlessly to meet growing demand while keeping costs sustainable. Instana was already running containerized workloads on Amazon Elastic Kubernetes Service (Amazon EKS)—which businesses can use to start, run, and scale Kubernetes without thinking about cluster management—to achieve operational efficiency without manual intervention, faster application deployment, and faster response times. In addition, the company uses ClickHouse, an open-source columnar database management system and early adopter of Graviton, for analytics and reporting. “We need to support the sheer volumes of data that we’re dealing with and our large number of customers with the scale of our backend regions,” says Bailey.

After exploring ways it could optimize its workloads using Arm and Graviton, Instana began using Graviton3-based Amazon EC2 instances. Specifically, Instana used Amazon EC2 C7g instances, R7g instances, and M7g instances—which are optimized for compute-intensive, memory-intensive, and general purpose workloads, respectively—to improve their workload efficiency by 35 percent and reduce costs by 18 percent.

At that stage, Instana wanted to further enhance responsiveness and cost-efficiency. “Performance isn’t just about throughput; it’s about throughput with the lowest resource use, because every dollar spent on infrastructure is a reduction in profit margin,” says Bailey. Staying competitive in the software-as-a-service market means providing a service at the lowest sustainable cost for customers. Instana began working toward this goal by adopting Graviton4-based instances, which offer even better performance and efficiency than Graviton3-based instances. In just 4 months, the team migrated a large portion of the existing compute workloads for its observability tool to Graviton4. It will also use Graviton for any new workloads on Amazon EC2 and on Amazon EKS.

With its architecture fully containerized and running on scalable Graviton4-based Amazon EC2 instances, Instana provides a near real-time streaming analytics solution that can dynamically scale to meet demand and manage spikes in use. In addition, the efficiency of Graviton means Instana can maintain sustainable price performance as adoption grows. “We’ve reduced our costs per unit of business by approximately one-third while at the same time expanding our regions because of the growth of our customers,” says Bailey.

By migrating from Graviton3- to Graviton4-based instances, Instana further improved efficiency for the same near real-time analytics workloads on large data stores, reducing CPU use by an additional 35–40 percent. As a result, the company has more headroom to handle occasional large traffic spikes, improving resilience and capacity. In addition, Instana can scale down the size of some of its data store clusters, reducing costs by up to 40 percent.

As Instana implemented Graviton4-based instances, it also set up benchmarking for every code change it makes so that it can measure the solution’s efficiency. “We do load testing of all our systems to look for bottlenecks and do regression tests across our AWS environment before we roll out changes to our customers,” says Bailey. The company uses a combination of fixed workload and saturation tests so that its engineers understand the exact limits of its scale. In a saturation test, Instana simply increases the load until the system reaches its limit. Meanwhile, the fixed-workload test uses a specific number of simulated transactions, machines, and customers to test efficiency.

The company received support from the AWS team throughout its implementation of Graviton. “The AWS Solutions Architects we communicated with put a lot of time and effort into understanding what we have deployed and how it works so that they could best advise us,” says Bailey. “They brought as much information to us as we brought questions to them.” Overall, the company will grow and scale while delivering fast, reliable performance from its observability tool.

Instana has improved performance, responsiveness, and cost-efficiency by using Graviton-based Amazon EC2 instances. “Not only are we far more efficient in terms of compute, but we’re also running faster queries than we previously could by using Graviton4-based instances,” says Bailey.

With half of its targeted workloads running on Graviton, the company plans to continue its rollout of these instances to further enhance its observability tool. “This has been a very successful story for us, and our intent is to adopt Graviton-based instances for 100 percent of our workloads,” says Bailey. “If you are using AWS, you should be looking at using Graviton-based instances.”