S3 Synchronous and Asynchronous Replication Intro

ℹ️ Info
Document purpose
This document outlines the key VAST Data replication configurations based on three dimensions: Replication mode (sync vs. async), user source (local vs. external provider), and permission model (bucket vs. identity policies). For each scenario, we provide a concise description, explain when and why it's relevant, and link to detailed configuration steps. The goal is to help users and field teams identify the right setup for each environment and accelerate deployment decisions.

Introduction to VAST Data Replication

VAST Data replication provides robust data protection and disaster recovery by enabling the creation of replica copies of file systems, S3 buckets, and databases across remote clusters. Replication can be configured in two modes — synchronous or asynchronous — each tailored to different operational and resilience needs.

Synchronous Replication

Synchronous replication ensures zero data loss (RPO = 0) by replicating every write operation immediately and atomically from the source cluster to the destination cluster. Only after the data is successfully written to both sides is an acknowledgment returned to the client. This mode is ideal for mission-critical workloads where consistency across sites must be guaranteed in real time. While the destination copy is always fully up to date, failover is a manual process — RPO is zero, but RTO is not, as service restoration time depends on operational procedures.

Asynchronous Replication

Asynchronous replication captures the state of data at scheduled points in time using snapshots, and transfers only the changes since the last snapshot to one or more destination clusters. This approach supports flexibility and scalability, including many-to-one and one-to-many replication topologies, and is suitable for disaster recovery, long-term retention, and cross-region deployments. In the event of a failure, replicated data can be failed over and resumed, with a small risk of data loss limited to the replication interval.

Synchronous vs. Asynchronous Comparison Table

ℹ️ Info
Use this comparison table to choose the type of replication solution that best fits your operational needs.

Feature	Synchronous Replication	Asynchronous Replication
RPO (Recovery Point Objective)	Zero (no data loss)	Greater than zero (depends on snapshot schedule)
Write Behavior	Writes are replicated instantly before acknowledgment	Writes are acknowledged immediately; replicated later
Failover Readiness	Immediate, with fully up-to-date data	Near-ready; may lose recent writes
Latency Requirements	Low latency, typically <10ms	Works over high-latency links
Topology	One-to-one per protected path; for different protected paths, they can be multi-directional with multiple clusters.	One-to-one, one-to-many, many-to-one.
Use Case	Mission-critical, zero data loss	Disaster recovery, long-term retention
Write Access at Destination	Read-write under normal operation (read-only only if the secondary loses connection to the primary).	Read-only until failover.
Replication Frequency	Continuous (real-time)	Scheduled or on-demand

Bucket Policy vs. Identity Policy Comparison Table

Feature	Bucket Policy	Identity Policy
Scope of Application	Specific to a View (bucket)	Applied to a user or group, affects access across multiple views
Attachment	Attached only to a view	Attached to one or more users/groups
Replication Behavior	Replicated automatically with the view	Replicated manually; must be enabled and re-attached to the remote local users/groups
Lifecycle	Deleted automatically when the view/bucket is deleted	Persists independently of the bucket lifecycle
API Type	Managed through S3 API	Managed through VMS API
Best Use Case	When access rules are specific to a bucket/view	When should rules follow a user across many buckets/views
Multi-tenancy Suitability	Suited for view-based isolation	Suited for user-based identity control
External Identity Provider	Not directly integrated	No need to attach manually if using external identity provider