Ensure data correctness across multiple

nodes in distributed systems

Key Purpose
Balance between consistency, availability,
and performance (CAP theorem).
A Consistency Model is a formal contract
between a distributed system's data store and
the processes accessing it. Balancing consistency, availability, and
partition tolerance (CAP theorem).

Once a process reads a value, it will

Introduction to Consistency Models
Key Challenges
never see an older value.
Synchronizing updates across geographically
Email systems ensuring older emails Specifies how updates to replicas are seen across the system. distributed replicas.
Example Monotonic Reads Focused on individual clients maintaining
remain visible after syncs. Defines acceptable behavior for reads and writes.
a consistent view of the data.
Ensures writes by a single
process are applied in order. (e.g., Amazon DynamoDB, Google
Client-Centric Consistency Models Cloud databases
Distributed version Spanner).
Exampl Monotonic Writes
control systems like Git
Guarantees that a process sees Use Cases Distributed file systems (e.g., NFS, HDFS).
its own updates in subsequent
reads.
Blogs immediately reflecting DNS systems, caching mechanisms, and collaborative tools
Example Read-Your-Writes
the author’s changes
Ensures that writes following a
read are applied to the most Provides a strong
recent value Advantages guarantee of correctness.
E-commerce platforms ensuring All reads return the result of the
Example Writes-Follow-Reads most recent write, regardless of Impractical in large-scale distributed systems due to
cart updates are consistent
where the write occurred. synchronization overhead.

Strict Consistency Disadvantages Requires global clocks or perfectly synchronized systems.

Example Stock trading systems enforcing

Long-lived, stable replicas strict transaction ordering.
Permanent Replicas
in server clusters. All processes see operations in the Easier to implement than
same order, but this order need not Advantages strict consistency
Focused on maintaining a global view of Allows some level of asynchrony
align with real-time.
Created for performance the data store across replicas.
improvement (e.g., CDN replicas). Server-Initiated Replicas Replica Types May lead to performance
Sequential Consistency Disadvantages
Data-Centric Consistency Models bottlenecks if writes are prioritized

Created temporarily based on client Example Shared memory systems ensuring

Replica Placement and Synchronization program order for threads.
requests (e.g., browser caches). Client-Initiated Replicas (Models without Synchronization)
Consistency Models in Distributed Systems These models define how the entire
Notify replicas when data is outdated. Invalidation Protocols system behaves globally and ensure Extends sequential consistency by ensuring
consistency of shared data Advantages Guarantees correct operation ordering
real-time ordering of operations.
Updates applied to all Synchronization Mechanisms
Active Replication High performance cost due to
replicas simultaneously. Linearizability (Atomic Consistency) Disadvantages
real-time constraints
Updates sent automatically by the server. Push Example: Banking systems ensuring debits
Push/Pull Protocols occur before withdrawals.
Clients request updates when needed. Pull
Guarantees that causally related
operations are seen in the same
order by all processes. Operations not causally related may be
Concurrent Operations observed in different orders.
Causal Consistency
Social media: Comments must appear
Example
after the post they reference.

Writes from a single process are

observed in order, but writes from
Suitable for large-scale systems where Ensures all replicas eventually different processes may be seen in
Advantages converge to the same state. varying orders.
immediate consistency isn’t critical. Advantages Lightweight implementation.
FIFO Consistency
Eventual Consistency (PRAM Consistency)
Temporary inconsistencies can occur Disadvantages Example Logs from distributed processes.

Uses synchronization variables

DNS resolving domain names. Example Ensures consistency only after explicit Mechanism to enforce consistency.
synchronization operations.
Weak Consistency
Focused on relaxing consistency Example Parallel programming environments with
Efficiently spread updates to all guarantees to improve shared memory
replicas using minimal messages. performance and scalability.
Acquire Ensures updates are visible
Example Epidemic Protocols before accessing data
Weak Consistency Models Uses acquire and release Definition
operations for synchronization.
Release Propagates changes upon
Randomly synchronize exiting a critical region.
(Models with Synchronization) Release Consistency
updates between replicas. Advantages Provides better performance by separating
synchronization from data access.
Anti-Entropy Protocols Advanced Protocols
Example Distributed file systems using
locks for access control.
Incrementally spread
updates among replicas. Synchronization applies at the
level of specific data items, not Advantages Efficient for localized critical sections.
Gossip Protocols
the entire data store.
Entry Consistency
Collaborative tools like Google Docs
Example
managing document sections