Ch 2: NoSQL Concepts

[[deleted user]]

I really enjoyed this chapter. It clearly and concisely compares and contrasts traditional relational database technology with the NoSQL family. I often see confusion arising in other comparisons because the SQL/relational database model is a single paradigm, and NoSQL is an umbrella term for many different paradigms, so there are many possible comparisons, not just one. This chapter makes it clear that NoSQL is a family that offers a spectrum of features and options.

On the CAP theorem topic, Hadoop picks C and P, although there is some debate over the applicability of the CAP theorem to distributed NoSQL systems. For example, see http://blog.cloudera.com/blog/2010/04...

[Aleksander]

Besides NoSql concepts described in this chapter I found the discussion on using consistent hashing interesting. For some reason I never realized that source control tools use hashing for synchronization.

This is first time when I find reading an electronic copy of the book more enjoying than actual book. I couldn't really tell what’s represented on black and white pictures, for example in Figure 2.8. ACID versus BASE, until I looked at the color pictures in the electronic copy.

[Dan]

Here are a few review questions you can ponder as you are reading chapter 2.

Unix pipes promoted small modular tools that each did one function well but could also be used with other components. What are some features of NoSQL systems that promote modular application architectures and modular databases?

How do tiered application architectures impact the role of NoSQL databases? What are some of the pros and cons of creating a REST data access layer between the application and the database?

What are the pros and cons of using an search function that is built into your database vs. using external service for search?

Many NoSQL systems and distributed file systems like HDFS are "rack aware" in that they can be configured to store a file on at least two different racks of servers. This is handy if one rack fails. What steps must be done to make a traditional SQL database "rack aware"?

What are typical read and write speeds for a flash drive or a solid state disk? How do these compare to a traditional rotating hard drive? How will fast-reads on SSDs impact the need to store indexes in RAM?

Let me know if these review questions are useful.