Join us at ScyllaDB University LIVE, instructor-led training sessions | March 19
Register now

Application Workflow And Query Analysis

2 min to complete

Application Workflow And Query Analysis

An early part of the data modeling is understanding how users will interact with the application. This allows us to understand which queries will be performed and how often. Remember, data modeling in ScyllaDB is query-based. We have to think about the application and the queries and take them into account when creating the data model. 

Going back to our 4Paw Veterinary Clinic example, let’s look at the queries.

Initially, an administrator logs into the system. She can add new pets and assign them a chip_id. After that, pet health information is recorded every five seconds.  A pet owner can request health information for a given pet on a given day. Once we have that workflow mapped out, we can think of specific queries, what data will be written/read, and how often. Some of the queries might be: 

  • Query 1: User logs in to the application
  • Query 2: Update a pet’s health information
  • Query 3: Read pet’s health information on a given day

In a real-world use case, we would have to develop these queries further and think about the conceptual data model (the entities and the relationships between them). 

Some important things to keep in mind:

  • Aim at creating a single partition per query. If a query needs to access only one partition, it would be very efficient. If multiple partitions need to be accessed for a single query, this can be acceptable, but it would be less efficient. If multiple partitions are being accessed for a query that’s being used often, we get less efficiency and maybe something in the data model is wrong.
  • Avoid scanning the entire cluster to find the data
  • Avoid scanning an entire table for the information needed (linear search).
fa-angle-up