Lightning talk: building a cloud of fares
0 likes54 views
Ralph Ligtenberg discusses moving Travix's "Fare Cache" system for storing airline fare search results from an on-premise Cassandra database to Google Cloud. The original system was difficult to scale and maintain. The new "Fare Cloud" uses microservices written in Go hosted on Google Cloud Platform services like BigTable, Datastore, and Dataflow. It took 6 months to implement and provides better performance and cost savings compared to the original system. The new architecture stores fare search results in the cloud to serve multiple use cases for Travix's products.
1 of 12
Download to read offline
More Related Content
What's hot (20)
Ad
Similar to Lightning talk: building a cloud of fares (20)
Ad
Recently uploaded (20)
Lightning talk: building a cloud of fares
- 1. Building a cloud of fares Ralph Ligtenberg February 21, 2019
- 2. Ralph Ligtenberg Tech Lead @ Travix .Net/C#, Go, Google Cloud Platform twitter.com/prutswonder travix.io/@prutswonder linkedin.com/in/ralphligtenberg Who am I?
- 4. ● Key principle: Store all search results → query → current cheapest price ● One product, multiple use cases ● Cassandra & .Net ● Released beginning of 2015 History of the “Fare Cache”
- 5. ● Scalable, though limited & expensive ○ Micro-monolith ○ Ran on a single machine! ● Cassandra requires advanced Ops knowledge ○ Not maintained by Cassandra expert ○ Tombstoning ○ Replication factor vs cluster size Good, but not future-proof
- 6. ● “One Cache To Rule Them All” ● Move data to Google Cloud ● Build horizontally ● Integrate with “old” solution ● New solution saves money! The cloud idea
- 7. ● BigTable, Datastore & Dataflow ● Micro-services in Go ● Only specific use-cases ● Implementation: 6 months Enter “Fare Cloud”
- 8. ● .Net Core v1.0 just released ● Most Google Cloud drivers were still in beta ● Easy to learn (hard to master) ● PoC was more performant than expected Why use “Go” instead of .Net Core?
- 12. Thank you!