How Coupa Reduced Their Elasticsearch Expenses by 60% and Improved Performance 

By Opster Team

Updated: Mar 10, 2024

| 3 min read

About Coupa

Coupa is a global technology platform for Business Spend Management. Their cloud platform allows users to manage all transactions across procurement, payments, and supply chain. By analyzing large quantities of corporate transactional expense data, the platform provides users with recommendations for improved inventory management, smarter purchasing decisions and lower costs.

Main takeaways

  • Reduced Elasticsearch costs by 60%
  • Optimized performance and eliminated issues
  • Improved capacity planning, upgrade strategy & more

Looking for better ways to support Coupa’s massive growth  

As a data-driven company, Elasticsearch has been one of Coupa’s key technologies for the last 5 years. “We use Elasticsearch for both of our primary use cases: our product search and logging,” said Arpit Ghiya, Senior Lead SRE. “Before we started working with Opster, we were always looking for better ways to manage the templates, to optimize the usage of the clusters, and to handle any incidents that occurred.”

Based on anticipated growth in traffic, the Coupa team began looking for a solution to help them handle the imminent scaling.

“There were a lot of factors we were considering. We wanted something that could provide us with proactive monitoring and alerting solutions, that would ensure platform stability, and would alert us to issues before they actually occurred and became problematic. However, we also wanted a solution that would help us reduce costs as we scaled.” Ghiya said.

When searching for possible solutions, the Coupa team came across Opster’s free Elasticsearch Check-Up. “We liked the recommendations that we got, and that’s when we engaged in our official process with the Opster team and got started with the AutoOps platform.”

“What we wanted was to ensure our engineers wouldn’t have to worry about platform stability at all, and with Opster we achieved that.”

Arpit Ghiya | SSenior Lead SRE

Arpit Ghiya | SSenior Lead SRE

The ultimate solution for ensuring performance & stability while reducing costs 

“For the first few months of our work with the Opster team and Opster products, our core focus was to ensure that our clusters were all stable and reliable. We found Opster’s AutoOps service to be a powerful technology that gave us alerting, analysis and resolution of incidents in real-time, within the SaaS. By working with the platform’s indications and recommendations, we improved performance a lot. We majorly reduced the number of incidents and alerts, and we continuously reviewed the cluster state and steadily optimized it,” Arpit Ghiya shared. “What we wanted was to ensure our engineers wouldn’t have to worry about platform stability at all, and we achieved that.”

“Since beginning to work with Opster, we’ve been able to reduce our hardware costs by 60% every month.”

Arpit Ghiya | SSenior Lead SRE

Arpit Ghiya | SSenior Lead SRE

Aside from cluster stability and performance, the major change that occurred after beginning to work with Opster was the cost of Coupa’s Elasticsearch operation. “Since beginning to work with Opster, we’ve been able to reduce our hardware costs by 60% every month. The massive reductions in cost are still ongoing, we’re still counting.”

During this time period, Coupa’s customer base continued to increase, so not only were costs reduced from what they were before, the cost would actually have continued to grow along with the growth in data. “Honestly, Opster saved us both infrastructure costs and the human cost of our team’s time and effort.”

“Opster saved us both infrastructure costs and the human cost of our team’s time and effort.”

Arpit Ghiya | SSenior Lead SRE

Arpit Ghiya | SSenior Lead SRE

Elasticsearch optimizations and performance improvement

The Coupa team has finished a major release since implementing all the changes recommended by Opster’s team and platform into production, “and the benefits are reaping in now,” said Ghiya. These changes included optimizations aimed at reducing infrastructure costs and conserving resources. After connecting to AutoOps, shard optimization was a priority: improving capacity planning and adjusting the number of primary and replica shards needed for ideal performance showed immediate results. 

By using Opster’s Template Optimizer tool, the Coupa team optimized field mappings throughout their deployment which enabled them to reduce storage, therefore reducing costs. The teams built a live dashboard with all of the info on every query running on the cluster to improve query visibility for the Coupa application team. By using this dashboard and the Search Log Analyzer, the teams were able to detect inefficient queries that could be improved and heavy queries that were causing service disruption, leading to search performance boosts. 

“Since all of these optimizations, the overall experience has been better for both our internal team and our customers. The AutoOps platform is great and we’ve had amazing support from the Opster team.”

Arpit Ghiya | SSenior Lead SRE

Arpit Ghiya | SSenior Lead SRE

The Opster team also helped Coupa review internal processes on the cluster like replacing data nodes, upgrades or rolling restarts to help the cluster be more robust and stable. “There were a lot of things we weren’t using, like index lifecycle management,” said Ghiya. “The way we were managing data within the cluster or rolling over indices wasn’t standardized, which affected the cluster state. All of these things added additional costs and impacted our operation’s performance and stability.”

“Since all of these optimizations, the overall experience has been better for both our internal team and our customers. The AutoOps platform is great and we’ve had amazing support from the Opster team,” Ghiya said. “We used to be concerned about adding new functionalities or adding more data to the system, and now we don’t worry about it at all. We know we can onboard more and more because our cluster, with Opster’s optimizations, can handle it. We have nothing to worry about with Elasticsearch anymore.”