【Distributed Architecture and Clustered Database】

Describe common types of database architecture along with their benefits.
Describe some techniques for managing data and optimizing performance.

Intro

• RDBMSs offer distributed architectures for critical or large-scale workloads
• shared disk allows parallel processing with mechanisms for high availability during server failures
• shared nothing employs replication or partitioning for optimized performance
• database replication involves copying changes from one database server to one or more replicas
• sharding involves placing partitions on separate nodes, facilitating increased parallel processing, and improved overall performance

Distributed architectures

• Helpful for large-scale workloads
• Support high availability or scalability
• Involve clusters of machines interconnected through a network
• Enhance:
  • Scalability
  • Fault tolerance
  • Performance

Types of database architecture

• Shared disk architecture

  • Multiple database servers processing workloads concurrently

  • Server establishes a connection to a shared storage

  • High-speed communication between servers

  • Advantages:

    • Ensure effective workload distribution
    • Ensures scalability as demand grows
    • Reroutes clients during server failures
    • Maintain high availability
• Shared nothing architecture
  • Utilize either replication or partitioning techniques
  • Allows distribution of client workloads
  • Promotes parallel processing
  • Enhance fault tolerance by rerouting clients
• Combination and specialized architecture
  • Employ a combination of:
    • Shared disk
    • Shared nothing
    • Replication or partitioning techniques
  • Integrate specialized hardware components

Help in managing data and optimizing performance

• Common techniques:

• Database replication
  - Involves copying changes to replicas
  - Distributes the client workload across servers

Replica resides in the same location, we call it a High Availability (HA) replica.
If the primary database server experiences a failure due to software or hardware issues, the system redirects clients to HA replica.
To mitigate broader disasters, organizations establish replicas in geographically distributed locations. This guarantees that during instances of complete data center outages, be it due to power loss, fire, earthquake or flood, clients can be rerouted to disaster recovery replicas.

• Database partitioning and sharding
  • Involves partitioning tables with substantial data
  • Each shard possesses its compute resources
  • Increased parallel processing and improved overall performance
  • Advantages:
    • Handle data warehousing and business intelligence
    • Handle extensive volumes of data