Thursday

Book Review: Cassandra High Availability

Offtopic: My first book "High performance in-memory computing with Apache Ignite" has been released and available at http://leanpub.com/ignite.

This post is my review of the Packt Publishing book Cassandra High Availability by Robbie Strickland. As the main title suggest. The book has almost 186 pages covering 9 chapters and must read for Cassandra users.

Chapter 1: Cassandra's approach to High Availability
The initial chapter of Cassandra High Availability cover some architectural design such as Monolithic Architecture, Master slave Architecture and the Cassandra's approach to achieve high availability. Most of all modern software system requires a non functional requirements such as High availability. In this chapter author briefly describe why RDBMS is the single point of failure and is not suitable for horizontal scaling, also the drawback of the Master Slave architecture. In the next few section Author describe the Cassandra's architecture and it's approach to solve the problem with high availability, also introduce Cassandra's replication mechanism which make Cassandra DB more unique than other NoSQL DB.
Chapter 2: Data Distribution
This chapter introduce Cassandra's Data distribution strategy such as consistency hashing, tokens, vnodes and various practitioner. Author briefly describe Cassandra token arrangements such as manually assigned tokens and Vnodes, even more how vnodes can improve availability, proc and cons of all the strategy.
Section partitioning discuss all the three options and difference between them, also point out when to use Murmur3Partiotion and the benefit of the ByteOrderedPartitioner. Understanding of these fundamentals will help readers to scale cluster effectively.
Chapter 3: Replication
Chapter provide fundamental concepts of replication and consistency of Cassandra cluster. Author demonstrate relation between replication and various consistency level and how they impact on Cassandra performance and data consistency. From my point of view very interesting part is the ,which can be very useful to make decision for deploy very wide range of cluster.
Chapter 4: DataCenter
Datacenter or DC one of the basic features for Cassandra high availability. Chapter provides all the possible use cases to use multiple data centers. One on the most important part of this chapter is the replication across data centers, where author describe the consistency in a multiple data centers. Also chapter covers how to setup multiple data centers and analysis data through Hadoop and spark.
Chapter 5: Scaling out
Chapter focused on scaling of Cassandra cluster, provide information for hardware sizing. Chapter base on all the valuable information of how to setup cluster, add nodes and decommissioning of node on the cluster. All of the administrative information's is very useful for any Cassandra administrator.
Chapter 6: Java Client
Mainly this chapter dedicated for developer and author provides all the necessary information to develop high available application based on Cassandra. Most of the code snippet based on Data stax native java client and graceful show how to connect to cluster, using asynchronous request and using failover policy. Chapter shows most of all the wealthy functionality of native client. For the Cassandra beginners current chapter might be the quick start point.
Chapter 7: Modeling for high availability
Chapter bases on the Cassandra Data model, author successfully describe the Cassandra's low level storage model. Author provides information every level of compaction and how compaction works under the hood. Even more, explain when and how to use different level of compactation to achieve desired goal. End of the chapter demonstrate how to model time series data and how to work with geospatial data.
Chapter 8: Anitpaterns
To be hones, it's best chapter of this book, which shows why we should not use secondary index and why we shouldn't use Cassandra as a distributed queue. Chapter demonstrate how tomstoms can effect on Cassandra read performace and how to get rid out of it. This chapter can give readers a foundation that will allow to make correct data modeling decision.
Chapter 9: Falling gracefully
This chapter mainly aim to the Cassandra administration and give well defined information for monitoring and management Cassandra cluster. Chapter clearly define all the possible way to monitoring Cassandra, such as through JMX , nodetools or third party tools. End of the part describes backup and restore process of Cassandra which will be very useful for Cassandra administrator.