| Author: | Rusty Klophaus |
|---|
- Dynamo-inspired KV, consistent & easy
- NoSQL
- More companies, funding, awareness
- Comcast, Wikia, Joyent (Riak customers)
- Why?
- Driven by narcissism, voyeurism & materialism
- Data generated by humans
- Constantly connected devices
- Social Networks (connections that make no sense to computers but make perfect sense to humans)
- Blob-y, Schema-less, High availability, Capacity
- Turning Point
- Scratch an itch -> Build a company
- Four Important Things
- Broad Knowledge (Don't make the mistakes of the past)
- Strong Community (Real problems)
- Agility (Solve new problems)
- Restraint (No solution can do everything well)
- What Characteristics of Riak Important After Different Cluster Sizes?
- Single Box Riak
- Might as well use a regular DB
- Simple KV, Flex Schema
- Clients in Many Languages
- Predictable Dev Interface
- Scales up & down well
- Same interface as a large cluster
- Configurable buckets (like tables-ish)
- Ability to change replication per bucket, write to different drives, etc.
- Links (Lightweight Relations)
- Small Riak Cluster
- Parallelism
- Expand the cluster
- Distributed queries (but act like one)
- Javascript Map/Reduce
- Well-Behaved HTTP
- Full-Text Search
- Power
- Large Riak Cluster
- Operations
- No Special Nodes
- Laugh at machine failure (Replication & Vector Clocks)
- Scale by adding nodes
- Packages or Self-Contained Installation
- Plug OSS & Enterprise for 10's of boxes
- Better off non-virtualized