EveryMundo provides Offer Marketing Software to drive direct channel growth and customer acquisition for major travel companies worldwide. The company works with over 60 airlines worldwide ranging from major full-service carriers to LCCs, based on 6 continents and marketing in over 25 languages.
Their products increase traffic, improve conversion rates, enhance user experience, and drive brand engagement on and off the airline’s website.
Operating Elasticsearch With a Managed Service
EveryMundo operates a mission-critical Elasticsearch deployment. Their team was constantly engaged in keeping their clusters up and running, despite the fact that they were using Elastic Cloud, which was purchased to provide them with a managed service. Their days often went to time-consuming Elasticsearch tasks, such as trying to troubleshoot issues and tune their cluster performance.
“It was a lot of back and forth with Elastic through a ticketing system,” said Jose Rafaelly, EveryMundo’s Head of System Engineering. “It’s one of those big companies where you have some sort of support, but there’s no face-to-face communication where they can help you right away, on the go unless you pay more. You open a ticket and at some point they’ll reply.”
Ultimately, Jose chose to leave Elastic Cloud because of the pricing relative to the value provided. “We were spending a lot of money on our cluster. I realized it would be more cost-effective to run the Elasticsearch clusters on our own.”
While looking for a monitoring tool to help them with their new deployment, Jose found Opster’s website. “It was a miracle. I believe it was.”
I realized for less than what I paid Elastic I could hold the cluster on my end and also afford to have Opster monitoring and troubleshooting all the issues.Jose Rafaelly, EveryMundo’s Head of System Engineering
Operating Elasticsearch Self-Hosted With Opster
“The first thing I did was try the free Elasticsearch Check-Up tool,” Jose said, describing the day he first discovered Opster’s website. A few hours later, he spoke with Opster’s experts and began a POC. “I realized for the same amount I paid Elastic I could hold the cluster on my end and also afford to have Opster monitoring and troubleshooting all the issues.”
“Since signing up with Opster, we haven’t had any issues in production. Dev is like production for us, and they help us with any issues we have there. They jump on and stay with us until everything is fixed. That’s been incredibly helpful, and something I really appreciate from Opster: to have that kind of support right away, not hours later.”
After troubleshooting all of the existing issues EveryMundo was dealing with in Elasticsearch, they began to work with Opster’s tools and solutions to optimize their system, starting with their memory heap. Throughout the process, all of EveryMundo’s client information was fully protected and was not accessed by Opster’s team, as both Opster and EveryMundo highly value protecting customers’ data privacy and upholding the highest security standards.
EveryMundo used Opster’s services and products to review instance types and how much heap was being used for Elasticsearch JVM. They set up the correct memory requests and limits in Kubernetes, covering all of the recommended settings and best practices. Garbage collection was tuned specifically according to EveryMundo’s use-case. Throughout the process, the EveryMundo team ran many benchmark tests in order to validate the optimized settings.
As part of the general data review conducted with EveryMundo, templates and mappings were optimized. EveryMundo’s benchmark tests validated a 50% improvement in search and index performance, just from the data schema tuning alone.
By optimizing data templates and mappings, search and indexing performance improved by 50%. Using Opster’s Operator for Shard Optimization, search throughput performance improved by 30%.
Using Opster’s Operator for Shard Optimization, many small indices were merged into optimized and larger shards. This process reduced the amount of shards in the cluster by hundreds and improved search throughput performance by 30%. Bottlenecks and hotspots were reduced throughout the cluster, resulting in a balanced load across all the data nodes.
“We’ve been learning through the process as well,” Jose said. “That’s important too.”
Thanks to all of these optimizations and Opster’s ongoing tuning, “We’ve been able to not think about Elasticsearch anymore. We can focus on other aspects of our infrastructure and let Opster take over the Elasticsearch work.”
We’ve been able to not think about Elasticsearch anymore. We can focus on other aspects of our infrastructure.
“Every meeting that we have, I thank the Opster team for all the help Opster provides. I’m very glad we started this partnership and can continue working together in the future. I believe it’s a team with a lot of experience who have made great solutions for managing Elasticsearch. Opster is very professional, and as people they’re awesome – so it’s the entire package. If we keep getting the support we’re getting right now, we’re good for years to come.”