This presentation tackles a particularly challenging situation that often occurs when creating a distributed relational database. In this presentation you will learn: - What a ‘shard conflict’ is - How to identify ‘shard conflicts’ - How to resolve ‘shard conflicts’ in a distributed database - How ‘shard conflicts’ affect query processing
Text of Database Scalability - The Shard Conflict
Database Scalability: The Shard Conflict
The Database Scalability: The Shard Conflict
This presentation tackles a particularly challenging situation that often occurs when
creating a distributed database.
In this presentation you will learn: • What a ‘shard conflict’ is• How to identify ‘shard conflicts’ • How to resolve ‘shard conflicts’ in a distributed database• How ‘shard conflicts’ affect query processing
Traditional Databases vs. Distributed Databases
Traditional Monolithic DBMade up of tables of data that are
related to one another
Modern Distributed DBData distribution is necessary for
All of the data is located in one place and is easily accessible
Information is spread across various servers (instances)
The data relationship is stored deep in the database and can be easily analyzed and queried using conventional methods
Related data can be distributed into different partitions, or shards, making
related query requests difficult to process
So, What Is a‘Shard Conflict’?
At ScaleBase, we have coined the term ‘shard conflict’ to describe a situation where:
• A given statement cannot be executed as is, unchanged, on all (or one) partitions and cannot be relied upon to yield a truly correct result.
Let’s take a look at the following examples…
Identifying the Conflict
Choosing ‘id’ as the shard key presents a shard conflict, because there is no guarantee that all employees are in the same shard as their corresponding departments.
Resolving the Conflict
The Method• Choose
‘department_id’ as the ‘Employee Table’shard key
The Outcome:• The join query was
optimized as a result of all department-related data being stored in the same partition
• No cross-joins exist between partitions
• Statements can now safely be executed on all partitions
Wait a Minute...There’s Still a Conflict
‘Select e.first_name, e.last_name, m.first_name, m.last_name from employee e join employee m on e.manager_id=m.id’
Join the ‘Employee Table’ together with itself to find a manager there is no guarantee they are in the same shard.
The employee tables are not capable of being sharded by both ‘id’ and ‘manager_id’ at the same time.
‘Shard Conflict’ Effects on Query Processing
• It is clear from the examples that when dealing with a foreign key and two tables, a common key can be utilized to resolve certain (but not all) conflicts
• Distributed data can become quite complex if not handled correctly
• It’s the kind of problem that is not always obvious, and can yield incorrect results, unnoticed
ScaleBase Can Help
ScaleBase is a modern, distributed MySQL database management system. It is optimized for the cloud and deploys in minutes to enable you to scale out to an unlimited number of users, data and transactions. It is a horizontally scalable database cluster built on MySQL that dynamically optimizes workloads and availability by logically distributing data across public, private and geo-distributed clouds.