Backend at trivago

The price of reliability is the pursuit of the utmost simplicity.
— C.A.R. Hoare, Turing Award lecture

On a normal day, we ingest a lot of data into our ELK clusters (~6TB across all of our data centers). This is mostly operational data (logs) from different components in our infrastructure. This data ranges from purely technical info (logs from our services) to data about which pages our users are loading (intersection between business and technical data).

Imagine you go to your hotel for check-in and they say that your dog is not allowed even though the website clearly states that it is!

To the outside, trivago appears to be one single software product providing our popular hotel meta search. Behind the scenes, however, it is home to dozens of projects and tools to support it. Teams are encouraged to choose the programming languages and frameworks that will get the job done best. Only few restrictions are placed on the teams in these decisions, primarily long-term maintainability. As a result, trivago has a largely polyglot code base that fosters creativity and diverse thinking. It allows us to make informed decisions based on actual requirements rather than legacy code or antiquated projects.

Make was created in 1976 by Stuart Feldman at Bell Labs to help build C programs. But how can this 40+ year old piece of software help us develop and maintain our ever-growing amount of cloud-based microservices?

trivago Intelligence was born in 2013 with two main objectives: First, to provide bidding capability to the advertisers, who are listed on trivago, and second, to provide them with metrics related to their own hotels; like clicks, revenue, and bookings (typical BI data). This project faced a wave of inevitable data growth which lead to a refactoring process which produced a lot of learnings for the team. As I expect it to be useful for other teams who deal with similar challenges, this article will describe why a team started a full migration of technologies, how we did it and the result of it.

Adopting an automation-first mindset is the first step to reduce manual and repetitive work. Thinking this way enables us to move faster, and more efficiently. It unburdens us from mundane, repetitive work, allowing us to focus on solving problems and creating value in the Software Development Life Cycle.

Many of our data pipelines interact with external services. The availability of an external service can adversly affect the health our pipelines. This is how we handle it using AWS Step Functions

Hello from trivago’s performance & monitoring team. One important part of our job is to ship more than a terabyte of logs and system metrics per day, from various data sources into elasticsearch, several time series databases and other data sinks. We do so by reading most of the data from multiple Kafka clusters and processing them with nearly 100 Logstashes. Our clusters currently consists of ~30 machines running Debian 7 with bare-metal installations of the aforementioned services. This summer we decided to migrate all of this to an on-premise [Nomad](https://www.nomadproject.io/ cluster) cluster.