[izzle49]

this seems like an ad for cassandra, many comments on how it scales and how partitioning is easy but no concrete examples at all. Having used cassandra in the past, it sure isn't the holy grail of nosql dbs, especially not for analytics that this article suggests

[cmrdporcupine]

I have used Cassandra in the past and would consider using it again but I agree its strength is not for 'analytics'. I'd say its strength is needing to scale reads and writes out on large large quantities of relatively simple data that you do not want to lose and you need low latency access to, and for when data needed to be easily partitioned across machines and data centres. I used it for data needed in realtime bidding systems for ad exchanges. It seemed well suited for that.

Then again, it's been 5 years since I worked with it, there have been improvements. Back then it had a _lot_ of warts, and I spent a lot of time fighting JVM GC pauses.

[throwaway859876]

Warts are still there, maybe worse with vnodes. Enjoy!

[nimrody]

Can you expand on why Cassandra is not good for analytics? Is that because you have to think of the queries before designing the data model? Or perhaps there are other issues?

[izzle49]

it isn't very good for aggregation of data. It doesnt provide core functionality that you need for analytics such as grouping, joining etc. The way to achieve this is to know what analysis you want to do before you store the data, which isnt practical for different types of analysis on the same dataset