Motivations

From developers

• Handling data is hard
• “I don’t want to create another table just to store this” should not be an excuse, I should be able to create tables whenever I want
• handling Protobufs is easier than a SQL table and I can reuse it for serde operations
• key-value stores are great, but it takes time to create value on top of it
• I need transactions to be consistent over multiples writes
• we should use something with strong capabilities:
  • exhibit A from MongoDB
  • exhibit B from ETCD
  • or any Jepsen’s analysis

From operators

• Handling data is hard,
• non-distributed databases are hard to manage,
• distributed systems are also hard but offer greater capabilities,
• when we have a massive distributed key-value store, we should be able to dedicate a keyspace to a user and a context,