Concurrency / Que with clustering

I realize a lot of people depends on existing framework or infrastructure to achieve concurrency or Que system to achieve what they need.

Basically if one need to get multiple server to be synced for write, and read, its possible to do it manually by coding it with combination of python or php scripts to connect between nodes...

But faster way to achieve so is by using open source combination such as zookerper with cassandra and cage.

But Case have issue when you are using more than 3 nodes.

This issue with zookeeper having to sync all the nodes together to have the transaction to run.

But after reading an article, its nice to have 2 or 3 nodes for transaction of 1 group of zookeeper process and separated zookeeper for read process. But do we need zookeeper for all nodes?

I think its better to have cassandra to sync those data to those server passively instead...

Only maintain between 2-3 nodes for transaction update to ensure uptime.