Samuel Sutch: CTO at Lightt
What does Lightt do?
Lightt is a visual timeline service that allows you to capture moments on your mobile device which are stored in an infinite visual timeline. Hard to explain, easy to understand once you see it.
How are you using Cassandra?
We are using Cassandra to store all of our online, live query data. Unlike other social services, we give users access to their entire timelines all the time, including dynamic ones like tags, your following timeline, etc, and cassandra is one of the tools that we use to enable that.
Why did you choose Cassandra?
Cassandra was chosen because of the simplified operations (things like partitioning and replication are handled for you) and simplified scaling it offers. I’m especially excited to migrate to 1.2 (excited about using cassandra post-migration, not about the migration itself because it offers us a way to even more flexibly resize the cluster. Other reasons certainly apply: flexible schema for object storage, sorted column model works great for timeseries (like our timelines).
Do you have any advice for people just getting started with C*?
Create a schema, seriously. Consider how migrations might have to work. Carefully consider how you serialize complex structures into columns (just use json). Read ALL of the docs: how consistency in C* works, ops (especially move token, although I guess only really valid if you are not using vnodes, post 1.2), join and repair. Your cassandra cluster is not write-read. Use counter column types for counting things. Do test runs on upgrades.
Are you running C* in the cloud or your own data center?
On AWS using the DataStax image from 1.0.
Anything else you’d like to add?
I would love to see some open source graphical tools pop up and be publicized by DataStax. Like RethinkDB, their graphical admin tool is amazing, and really help one get their head around how the cluster is laid out and operating.