Introduction
Azure Cosmos DB, a multi-model database service from Microsoft, is lauded for its offering of global distribution and horizontal scalability. A noteworthy feature of Cosmos DB is its provision of five unique consistency models.
Data consistency in a database focuses on ensuring that reads reflect the writes. As data is written or updated, these changes need to be reflected accurately when subsequently read.
The Consistency Spectrum
Azure Cosmos DB offers five types of consistency levels placed on a spectrum, starting from "strong" to "eventual."
1. Strong consistency
The "strong" model is the gold standard of data consistency: a read always reflects the most recent committed write. But, this consistency comes at the expense of availability and latency during network failures.
2. Bounded staleness consistency
This model provides a middle ground by ensuring consistency within either a specified lag time or a given number of write operations. Meaning, any read operations will be lagging at most 'k' versions (number of operations) of the item or 't' interval units (time interval) behind the write.
3. Session Consistency
Session consistency ensures that within a single session, the client always reads the latest writes made by that client. It provides a natural model for web applications providing monotonic reads, monotonic writes, and read-your-own-writes functionalities.
4. Consistent Prefix (Ordered) consistency
Here, the read operations return the subset of the writes in the same order. Consistent prefix guarantees that reads never observe out-of-order writes.
5. Eventual consistency
On the other end of the spectrum is "eventual" consistency, the lowest consistency and highest availability model. In this model, a write will eventually be seen by subsequent reads in the absence of any further writes.
![Eventual]()
Understanding consistency levels with a shopping cart application
To further comprehend these concepts, let's use the example of an online shopping cart application.
1. Strong consistency use case
Consider a hot-selling item on the platform. A customer decides to buy it. With strong consistency, the system immediately reflects this purchase to all other users, decreasing the available inventory count by one. So, no other customers end up buying an already sold out item, ensuring a smooth, accurate inventory management.
2. Bounded staleness consistency use case
Suppose customers write product reviews. With bounded staleness, newly submitted reviews would not appear for others immediately. However, customers can be assured that they will see the new reviews within the delay defined by the 'k' operations or 't' time interval.
3. Session consistency use case
Let's say a user shopping across multiple devices adds an item to their cart. With session consistency, they can continue their shopping spree from any device, and the added item will be immediately visible in their cart, ensuring a seamless user experience.
4. Consistent Prefix (Ordered) consistency use case
Imagine a massive holiday sale where thousands of users are adding different items to their carts simultaneously. With consistent prefix consistency, the system registers the products being added in sequences, ensuring no sequences are skipped or shuffled – vital for maintaining the right order of operations.
5. Eventual consistency use case
In an online store, showing the trending products based on most added to the cart can use eventual consistency. It allows the application to continue functioning smoothly while the "trending items" list takes some time to reflect the changes.
Conclusion
Opting for the correct consistency level is no one-size-fits-all situation. It varies per the application's needs, ensuring the right balance of consistency, latency, and availability. Azure Cosmos DB's flexibility further empowers developers to fine-tune their apps adjusting consistency models as the application evolves. A well-informed decision based on understanding each consistency level's nuances ensures an alignment with the application design and an optimal end-user experience. For a deeper understanding of Cosmos DB's consistency levels, please refer to the official Microsoft Azure documentation.