Download case study

Get Started

Fill in your details and one of our experts will reach out shortly.

Thank you

Thank you for reaching out.
We will be in touch soon. Trims GKE Travel Tech Workload Core Count by Over 20%

Download Case Study
Reduced CPU Utilization
Week for Results trims GKE travel tech workload core count by over 20% without any R&D efforts and within a single week with Granulate

About is a leading travel tech company serving global customers and headquartered in the Czech Republic. It employs over 1,000 people worldwide who work to bring its innovative Virtual Interlining algorithm to market, which allows users to combine flights across legacy and low-cost airlines into one single itinerary.

Granulate helped us improve performance and with a cost reduction too, cutting our CPU utilization by 20% and keeping latency steady
Alexandru Viscreanu, Backend Engineer

The Challenge

With servers in Google Cloud’s Kubernetes platform GKE, higher web traffic and utilization of’s flight search engine translated into more cores activated per request, and higher costs each month as the elasticity of its infrastructure responded to demand. Cluster expansion went to keeping the flight search service quick and accurate, with a low-latency experience for customers who were engaged by the service’s speedy delivery and results.’s growth as a technology service was matched by infrastructure growth in the short term, but without cost optimization this strategy would consume an inordinate amount of revenue. Performance optimization was a bonus: as long as latency was below a certain threshold the customer experience would remain excellent.

The Results

Granulate’s immediate impact on clusters was a decline in latency which contributed to an initial reduction of at least 25% in CPU utilization. Such a descent triggered mechanisms within the HPA, automatically leading to a smaller number of pods deployed on behalf of the service — and therefore, an automatic reduction in’s cloud bill.

After a month of Granulate being active on two main clusters, results were benchmarked and final cost reduction estimates were realized: a 21.5% reduction in cores per request for its flight search service.


Scroll to top
Skip to content