Achieving High Availability: Strategies for System Redundancy

High Availability

To avoid interrupting operations, systems must incorporate elements that prevent downtime and minimize recovery time if an interruption occurs.

Required Elements

• Technology with a low failure rate
• Disaster tolerance (redundancy)
• Contingency procedures
• Switching mechanisms (manual or automatic)
• Change tracking

General Scheme of Redundancy

Layers in a Protected System

• Presentation: This layer represents the graphical user interface accessed by operators and customers via a web browser, served through a web server that manages context and sessions.
• Business Logic: This layer contains all business logic, handling transactions and consultations with local and external databases through server-side scripting languages.
• Data: This layer contains the physical data store and a database manager.

Data Replication

Data replication involves transporting data between two or more servers, allowing specific database information to be stored in multiple sites. This increases data availability and improves overall query performance.

Components

• Publisher: A server that makes data available to other servers for replication.
• Distributor: A database server hosting the distribution, storing historical data and transaction metadata.
• Subscribers: Servers that receive replicated data.

Replication Types

• Snapshots: A system where data is replicated as it is modified at the source.
• Transactional: A system for order processing and distribution. Multiple publishers can take orders, which are then replicated to a central warehouse. The store processes data in read-only mode and requires regular updates.
• Merge: Allows multiple sites to work online or offline independently, later merging data modifications into a single, uniform result. Data is synchronized at scheduled times or upon request. For example, a regional criminal history database allows municipalities to maintain local copies without requiring a permanent connection to the central body.

Determinants

• Autonomy: Measures how well a site can operate while disconnected from the publisher database.
• Transactional Consistency: Determined by the need to execute all transactions on the server or simply respect their order.
• Latency: Refers to the time required to synchronize data copies. Does the data need to be 100% in sync?

Conclusion: Replication is highly effective for improving data availability. In extreme cases, a fully replicated distributed database ensures high availability, as the system continues to operate if at least one server is functional. The primary disadvantage is the high cost of maintaining consistency across all sites.