Skip to content

Choose a deployment pattern

For the complete documentation index see: llms.txt

All documentation pages available in markdown.

Aerospike Cloud offers three reference deployment patterns, each a combination of zone count and replication factor (RF). Choose one before you provision a cluster. The pattern determines how your cluster behaves when a node or zone fails and how much data you can lose.

Before you choose

Each pattern trades off two dimensions of resilience:

  • Data availability: whether your application can keep reading and writing the full dataset when a node, zone, or region fails.
  • Data durability: how much data you can lose when a failure happens, expressed as a recovery point objective (RPO).

To find the right pattern, start with your requirements:

  • How available does your data need to be when a node or zone fails?
  • How much recent data must survive a failure (your RPO)?
  • Is the data a cache you can rebuild from another system, or is Aerospike the system of record?
  • How many availability zones (AZs) does your application run in?

Plan for these constraints, too:

  • The cluster size must divide evenly by the zone count and by the replication factor. For example, a 3-zone cluster with RF 3 needs a multiple of 3 nodes.
  • By default, a cluster can have up to 15 nodes. Contact support for a higher limit.

See Limitations for the full list, and Supported regions and instance types for available instance types. You are responsible for the business continuity and disaster recovery (BCDR) procedures that match your requirements. See the Shared responsibility model.

Compare deployment patterns

PatternZonesReplication factorBackupsSurvives a node failureSurvives a zone failureBest for
Single zone without replication11Not availableNoNoEphemeral cache
Single zone with replication12 or 3OptionalYesNoHigh-availability cache
Multi-AZ2 or 32 or 3OptionalYesYesOperational datastore

Each pattern can use disk or memory storage. With disk storage, Aerospike runs on its hybrid memory architecture (HMA). You can keep an entire working set in the cluster instead of caching only a subset.

A higher replication factor raises availability and durability, at the cost of more storage and write overhead.

Deployment patterns

The following sections describe each pattern and when to use it.

Single zone without replication

  • Use when: the data is an ephemeral cache that you can rebuild from a source system, and you can tolerate intermittent unavailability. It’s also the lowest-cost pattern, which makes it a reference point for comparing costs.
  • Availability: the cluster keeps one copy of your data. Whether one node or the whole zone fails, affected data becomes unavailable. Your application must fall back to the source system and rehydrate the cache.
  • Durability: a node failure permanently loses that node’s data, because no second copy exists and no backup is configured. Recovery depends on your source system, not on Aerospike Cloud.

Single zone with replication

One availability zone, replication factor 2 or 3.

  • Use when: you need a high-availability cache that survives node failures and you do not need protection from a zone outage.
  • Availability: the cluster keeps multiple copies of every record in the same zone. If a node degrades or fails, Aerospike Cloud uses the replicas to keep your data available. Because all copies are in one zone, a zone outage makes all data in the cluster unavailable.
  • Durability: if a node fails, Aerospike Cloud repopulates the replacement node from a replica with no data loss.

Multi-AZ

Two or three availability zones, replication factor 2 or 3. Choose the zone count to match how your application is deployed, and the replication factor to match your resilience requirements and cost tolerance.

  • Use when: you run a mission-critical operational datastore that must stay available when a zone fails.
  • Availability and durability: the cluster tolerates node and zone failures without manual action. A cluster with two zones and RF 2 survives one zone failure. A cluster with three zones and RF 3 survives two simultaneous zone failures.

More zones raise the availability ceiling but add cross-zone data transfer costs. Aerospike Cloud configures each zone as a separate rack. See Advanced configuration.

Extend a pattern

You can layer the following options on top of a pattern:

  • Multi-region replication: adds regional resilience and improves regional performance for multi-region applications. Active-passive deployments fail over to a secondary region, and active-active deployments serve reads and writes from multiple regions. Both use cross-datacenter replication (XDR).
  • Backups: available for the single zone with replication and Multi-AZ patterns. Backups support disaster recovery and compliance requirements, or provide a lower-cost durability option when an RPO measured in hours is acceptable. See Backups for the schedule and restore process.

Next steps