Getting Started with CockroachDB by Kishen Das Kondabagilu Rajanna

Author:Kishen Das Kondabagilu Rajanna [Rajanna, Kishen Das Kondabagilu]
Language: eng
Format: epub
Tags: high availability, database, database management, database administration, cloud native database
Publisher: Packt Publishing
Published: 2022-03-18T00:00:00+00:00

ACID from CockroachDB’s perspective

In this section, we will go over how each of the ACID properties is implemented in CockroachDB and what guarantees they provides. Like in the previous section, we will start with atomicity.

Atomicity

As we learned in the first section, atomicity ensures that all the statements in a given transaction are executed as a single unit – that is, either all of them succeed or all of them fail. This condition should be guaranteed irrespective of machine, network, and memory failures. This is essential to make sure multiple queries don’t run into each other.

CockroachDB allows you to have ACID transactions that can span the entire cluster, touching multiple nodes and geographical locations. CockroachDB supports this using an atomic protocol called parallel commits.

In the previous chapter, we learned about transaction records and write intent. A transaction record keeps track of the current state of the transaction and is maintained in the range where the first write occurs. Whenever we are changing a value, they are not directly written to the storage layer. Instead, a value is written to an intermediate state known as write intent. Write intent acts an Multiversion concurrency control (MVCC) record, with a link to the transaction record. Write intent acts as a replicated lock, which houses a replicated provisional value.

Write intent has been shown in the following figure with a sample transaction with two writes:

Figure 3.3 – Transaction timeline without parallel commit Any transaction that comes across a write intent should also go through corresponding transaction records and, based on its state, decide how to treat the write intent. A commit flips the transaction record state to committed. Once the transaction is committed, write intents are cleaned up asynchronously.

Now, let’s take a look at how traditional atomic transactions worked in CockroachDB without parallel commits and later with parallel commits.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.