Search as a Service with Julien Lemoine

“You need to build more things yourself to be highly available, but one of the very good consequences of being bare metal is that the prices are very low compared to what you could get on the cloud provider.”

Engineers who want to add search to their application usually deploy Elasticsearch, or write their own search engine that uses TF-IDF. These solutions work well for large documents, but are less effective for large volumes of small records–which is how many modern web and mobile applications are structured.

In today’s show, Julien Lemoine discusses how his company Algolia thinks about search. Algolia is a search as a service company that gives developers an easier way to search on their websites and applications.

Questions

• What are the unsolved problems in search?
• What is TF-IDF, and how is it is used to rank search results?
• Why did you decide to set up your own servers instead of going to a cloud provider?
• Why is it useful to divide search into two different aspects?
• How do you stress-test your system?
• How did you respond to your recent data center connectivity issue?
• How do you build a company oriented for the long-term?

Links

• Algolia
• Elasticsearch
• tf-idf
• Bare-metal servers
• Multitenancy model
• GAE
• Chaosmonkey testing strategy
• ZooKeeper
• Raft
• On premise
• Julien on Twitter
  

Pranay