Skip to content

Resilience on Aerospike Cloud

For the complete documentation index see: llms.txt

All documentation pages available in markdown.

Resilience on Aerospike Cloud is a partnership between what the platform guarantees automatically and how you configure your database. It has two dimensions:

  • Data availability: whether your applications can keep reading and writing your data when a failure happens.
  • Data durability: whether acknowledged data is protected from loss when a failure happens.

To pick a configuration before you provision a cluster, see Choose a deployment pattern.

Shared responsibility for reliability

Three layers make up total resilience. Aerospike owns the first layer. Aerospike builds the second layer and you configure it to match your requirements. Your application owns the third layer.

LayerOwnerWhat it covers
Core platform reliabilityAerospike (not configurable)Redundant underlying infrastructure, automatic node failure detection and recovery, and patching and maintenance that require no action from you.
Reliability-enhancing capabilitiesSharedDeployment pattern, replication factor, zone and region footprint, active-active or active-passive replication, and backups.
Application designYouClient-side retry and failover handling, load testing, the recovery point objective (RPO) and recovery time objective (RTO) your business needs, failover testing before go-live, and disaster recovery and compliance planning.

For security and operational responsibilities, see the Shared responsibility model.

Compare how failures affect deployment patterns

The following tables show how each deployment pattern responds to a node, zone, or region failure. Each table shows whether your data stays available and whether you lose any acknowledged data.

Node failure

PatternData availabilityData durability
Single zone without replicationData on the failed node stays unavailable until your application falls back to the source system and rebuilds from that system of record.Data loss for records that were only on that node.
Single zone with replicationNo interruption.No data loss.
Multi-AZNo interruption.No data loss.

Zone failure

PatternData availabilityData durability
Single zone without replicationAll data in the zone is unavailable.Potential data loss if the zone doesn’t recover.
Single zone with replicationAll data in the zone is unavailable.Potential data loss if the zone doesn’t recover.
Multi-AZData stays available and is served from the surviving zones.No data loss, up to the number of zone failures the pattern tolerates.

Region failure

PatternData availabilityData durability
Any pattern without multi-region replicationUnavailable.No data loss, unless the region suffers an unrecoverable disaster.
Any pattern with multi-region replicationBrief interruption during application failover. RTO is measured in minutes and depends on your application.No data loss, unless the region suffers an unrecoverable disaster.

Increase resilience

You can raise resilience with four options. Each option strengthens data availability, data durability, or both.

Replication factor

Replication factor (RF) sets how many copies of each record your cluster keeps.

  • Improves: data availability and data durability. More copies must be lost at the same time before data becomes unavailable or is lost.
  • Use when: you need your data to survive node failures. RF 1 keeps one copy and offers no cluster-level protection against data loss or data unavailability. RF 2 and RF 3 keep additional copies, which Aerospike Cloud distributes across the cluster automatically.
  • Trade-off: additional storage and write overhead. Each copy is replicated across nodes and, with a larger zone footprint, across availability zones.

Zone footprint

Your cluster can span one, two, or three availability zones (AZs) in a region.

  • Improves: data availability. Zone footprint, together with replication factor, determines whether your data stays available during a zone failure. A cluster in a single zone loses all data availability if that zone goes down, no matter how many copies replication creates.
  • Use when: you need your data to stay available during a zone failure. Zone footprint and replication factor together set how many zone failures the cluster tolerates. Two zones with RF 2 keep all data available during one zone failure. Three zones with RF 3 keep all data available during two zone failures.
  • Trade-off: cross-zone data transfer costs. Choose a zone footprint that matches how your applications are deployed, which aligns zone failure handling between your application and database.

Multi-region replication

Multi-region replication uses cross-datacenter replication (XDR) to replicate data asynchronously between clusters in different regions.

  • Improves: data availability during a full regional outage.
  • Use when: you need your data to stay available if an entire region fails.
    • Active-passive: The primary cluster handles all writes and ships them to the secondary cluster, which acts as a hot standby and can also serve reads. If the primary cluster becomes unavailable, your applications fail over to the secondary cluster.
    • Active-active: Multiple regions serve reads and writes at the same time, and each cluster ships updates to the others.
  • Trade-off: recovery requires a failover step. Recovery time is measured in minutes and depends on your application’s regional failover strategy.

Backups

Backups are scheduled exports of your database.

  • Improves: data durability. Backups also protect against broader events, such as accidental deletion.
  • Use when: you need disaster recovery or must meet compliance requirements. Backups are also a lower-cost durability option when a recovery point objective (RPO) measured in hours is acceptable.
  • Limit: you can add backups to every deployment pattern except single zone without replication. Deleting a cluster also deletes its backups. See Backups for the schedule and restore process.

Design your application for resilience

Application design is entirely your responsibility. See the Application design row in the table under Shared responsibility for reliability for the full checklist. It includes:

  • Client-side retry and failover handling
  • Load testing
  • Defining the recovery point objective (RPO) and recovery time objective (RTO) your business needs
  • Testing failover behavior before go-live
  • Disaster recovery and compliance planning

Next steps