How Tipalti mastered Elasticsearch performance with AutoOps

When Tipalti first moved its Elasticsearch workloads to Elastic Cloud, it followed the standard playbook. "Our standard procedure before AutoOps was to monitor Elastic activity with dedicated monitoring clusters for each environment," Levy explains. While functional, this approach quickly revealed its limitations.

This setup involves dedicated monitoring clusters where a separate Elasticsearch cluster is installed for each environment, such as development or production, solely to collect monitoring data and ship logs and metrics. This requires configuring all main application clusters to send their performance metrics and logs to these dedicated monitoring clusters. The core challenge was a lack of deep, actionable insight. "To diagnose an issue, we had to spend a significant amount of time manually analyzing nodes, charts, and indices to understand the root cause and determine a solution," Levy notes.

For a small team juggling multiple critical platforms, this became a significant time sink. They built custom Grafana dashboards as a workaround, but this required extensive collaboration with the DevOps team and still fell short of providing the clarity they needed. The team was spending too much time reacting to issues rather than preventing them.

Tipalti needed a solution that could do a specific "job" — demystify Elasticsearch performance and give the team back their most valuable resource: time. They found their solution in AutoOps.

The impact was immediate. 

"AutoOps is highly intuitive. The moment you open the platform, the overview provides immediate, comprehensive visibility into the state of your deployment."

Beyond the initial clarity, it was the actionable intelligence that proved to be a game changer.

"The recommendations feature is exceptionally powerful. It doesn't just identify a problem; it provides a clear explanation and the exact commands needed to resolve it. This is incredibly efficient and has been a significant time saver for our team," says Levy.

This was a fundamental shift. The team no longer had to hunt for answers; the answers were being delivered to them complete with solutions.

Adopting AutoOps translated directly into tangible benefits across Tipalti.

The most immediate financial win was decommissioning Tipalti’s old infrastructure. "We were able to decommission our legacy monitoring clusters entirely," Levy states. But the savings didn't stop there. With the deep visibility AutoOps provided into Tipalti’s indices and shards, the team identified a major optimization opportunity. They rearchitected their indexing strategy, moving from a costly index-per-customer model to a consolidated primary index with dedicated indices for only their top 10 clients. "By rearchitecting our indexing strategy, we achieved a significant reduction in both CPU and RAM usage, which allowed us to downsize our clusters," Levy explains.

AutoOps armed the team with the tools to get ahead of problems. Levy recalls an instance where "AutoOps proactively alerted us that a cluster's disk was nearing capacity, which enabled us to react and increase the disk size before it could cause any service impact."

Furthermore, the machine learning-powered anomaly detection capability has been crucial in reducing alert fatigue. By learning the natural ebbs and flows of the team’s workload, AutoOps avoids firing alerts during expected high-traffic periods. "The machine learning-based alerts are highly effective at eliminating false positives, which ensures our team only focuses on genuine issues," Levy says.

One of the most significant changes was cultural. Tipalti extended AutoOps access to its development teams, transforming the dynamic between infrastructure and application development.

This move democratized performance data, empowering developers to see the real-time impact of their code and take ownership of their services' efficiency.

For Tipalti, the journey with AutoOps has been one of empowerment. It has enabled a small, highly skilled team to effectively manage a critical, large-scale data platform, shifting their focus from reactive firefighting to proactive, data-driven optimization. They've cut costs, improved reliability, and built a more collaborative and efficient engineering culture.

Through strategic initiatives, the company has achieved a remarkable 10% annual cost saving, demonstrating a strong commitment to financial efficiency. Beyond monetary benefits, these efforts have also led to a significant improvement in system reliability, ensuring more consistent and dependable operations. Furthermore, a crucial outcome has been the cultivation of a more collaborative and efficient engineering culture. This transformation fosters greater teamwork, streamlined workflows, and an environment where innovation can thrive, ultimately leading to enhanced productivity and problem-solving capabilities within the engineering department.

Learn more about how Elastic customers use the Search AI Platform to achieve better business outcomes.