Request for Transactional Scenarios

[Henrique Moniz <hm...@gmail.com>]

Hi,

>From time to time discussions pop up here regarding the transactional or
atomic capabilities of Cassandra (or lack thereof). There is at least one
project dedicated to solving this problem (i.e., Cages). Unfortunately, in
pretty much every discussion or blog post I’ve come across on this subject,
people have not been very clear on stating their own reasons for requiring
transactional semantics.

I would like to hear from people who, at some point, felt they needed
transactional semantics in Cassandra. What motivated this need and did you
do about it? Did you come up with your solution or used Cages? And how did
that worked out for you?

I ask this because I’m very interested in contributing with a general
solution to this problem, but first it’s important to understand its extent.

It would be awesome if you could share whatever you can from your experience
(e.g., the kind of data, the higher-level operations performed on this data,
level of concurrency, throughput requirements, etc). Even if you feel you
cannot provide a detailed description, it would also be very useful if you
could at least describe the transactional patterns involved (e.g., number of
servers accessed, ratio of read/write operations, conflict rate, data flows,
etc).

Take care,
Henrique