Business overview

Our team, in collaboration with the Platform Engineering team of the client, helped to reduce the number of API requests to be processed for one of the main APIs of the client. We evolved and improved a caching solution originally implemented by the client in Go language, turning it into a scalable system that optimizes the network traffic and reduces time taken in processing and calculations. Our solution reduced the cloud infrastructure costs by optimizing the searches made in the search engine, consuming fewer resources. As a result, they had fewer cloud instances and saved about 15% of the current cost. 

Background

WebBeds is one of the few global bed banks that makes ground travel products accessible to more than 430000 hotels in more than 16000 destinations worldwide.

They offer powerful distribution solutions to hotels, making selling and buying accommodation easier. Thousands of travel agencies regularly connect to the WebBeds Marketplace and generate several billion searches daily.

Business need

The XML API service that WebBeds provides to its multiple stakeholders receives around 4 million search requests per minute, many of which are the same even in a brief period.

Caching the requests and their responses over a certain time was a key objective to allow fast responses for duplicated ones (a hit in the cache) and avoid processing them by the hotel search engine. This part, reducing the processing of duplicate searches, could significantly reduce cloud infrastructure costs (network traffic and CPU processing). Achieving a conservative hit ratio between 15 and 20% can directly reduce the same proportion of cost in the monthly invoice from the cloud provider. When we talk about a platform with around 4 billion requests per day, it is not a negligible figure.

Fast time of response
(less than one millisecond)  

Able to manage a high volume of requests (several millions per minute)

Adjusted infrastructure costs, the solution is saving much more than its own cost