Founded in 2014 in Berlin, Germany, the company operates local apps and websites in 25 countries across Europe, North America, South America, Australia, and Asia-Pacific. HomeToGo also operates brands such as Agriturismo.it, AMIVAC, Casamundo, CaseVacanza.it, e-domizil, EscapadaRural, Kurzurlaub, Kurz Mal Weg, Tripping.com and Wimdu.
The problem
The overall challenge was how to scale search support for 50 million visitors a month.
HomeToGo has more than 15 million accommodation offers on its platform, more than any other marketplace in the vacation rental industry. Real-time connections are essential for the company to work efficiently with suppliers’ APIs to check whether the accommodation is still available for the given dates, the specified number of travellers, and at what price. But it also needed to deliver HomeToGo’s fast and reliable market-leading service, which depends on a myriad of smart tools, ML, AI, trusted checkout, and payment solutions.
The entire mix is supported by advanced search functionality. In 2023, HomeToGo also launched its innovative HomeToGo Modes, a revolutionary new way to find vacation rentals and travel destinations, with its inaugural AI Mode. HomeToGo’s AI Mode enables guests to find their perfect vacation rental by describing it via written conversation to a generative AI-powered travel planner. With around 50 million monthly visitors during peak season, scalability and efficient search functionality had become an interesting engineering challenge to solve.
HomeToGo offers users powerful search capabilities, helping visitors to find their ideal vacation home at the best possible price. This functionality is underpinned by Elasticsearch, a critical part of HomeToGo’s operations. However, managing the world’s largest selection of vacation rentals while accommodating a rapidly growing user base proved operationally complex with a technology primarily designed for search use cases. Challenges included managing hundreds of servers and scaling the infrastructure multiple times daily without straining system management.
The solution
The solution lay in improving search functionality efficiency through architectural transformation.
To alleviate these operational burdens and improve the user experience, HomeToGo’s engineers opted to decouple price and availability data storage from search functionality, continuing to use Elasticsearch for search while adopting a different technology for price data storage. The chosen database needed to be fast, reliable and to seamlessly integrate with Elasticsearch. With these requirements in mind, HomeToGo selected Aerospike.
The implemented solution employs Aerospike to manage data provided by HomeToGo’s suppliers and to cache query results in Elasticsearch for repeated searches. This approach has significantly reduced the load on the Elasticsearch infrastructure by decreasing both the data volume stored in Elasticsearch and the workload, as cached data is returned for similar queries.
The result
The Aerospike implementation resulted in HomeToGo achieving sustainability, scalability, and improved cloud economics
HomeToGo immediately noticed that it could not only easily support customer traffic levels, but that the efficiency metric (the amount of servers needed to manage the search traffic) improved, while week-on-week server hours decreased.
- Improved cloud consumption – Between January and November 2023, HomeToGo experienced 37% fewer server hours consumed.
- Optimised storage cluster size – To support the high traffic numbers during peak season, HomeToGo only needs eight Aerospike servers, which drops to six in off-peak.
- Reduced Elasticsearch cluster size – Integrating Aerospike reduced the Elasticsearch cluster size by 50%.
One additional benefit of the migration to Aerospike was its ability to handle new hardware resources, such as the next generation of AWS instance types, which ultimately improved price and availability response times.
Audrius Bugas, technology director – Architecture, HomeToGo, said: “The travel business is fundamentally about price and availability, which needs to be fast and reliable. Aerospike’s ability to operate very well under pressure while improving our search response times has made it an invaluable partner in the future growth of our business. And with the significant reduction that we have experienced in server hours…that’s sustainability.”
Martin James, VP EMEA, Aerospike, said: “The task for Aerospike was to help HomeToGo to achieve real-time pricing and to ensure the availability of vacation rentals was always up to date without hitting a ceiling on scalability. The company had ambitious plans to grow, so our system, which supports AI/ML including generative AI – essential for HomeToGo’s AI Mode travel planner – is built for the future. In addition, with a considerable cut in its server numbers and server hours, the company has vastly improved its carbon footprint.”
