High availability architecture for HSTS

Learn how HSTS high availability clusters provide continuous enterprise-grade file transfer with automatic Redis-based failover.

Use a high availability cluster to ensure uninterrupted file transfers in enterprise environments where outages, node failures, or maintenance events are unacceptable.

An HSTS high availability cluster uses an active-passive architecture with Redis Sentinel to automatically fail over file transfers during node outages.

Cluster architecture

A cluster consists of nodes. A node is a machine (physical, virtual, or container-based) on which IBM Aspera High-Speed Transfer Server and Redis are installed. For Redis, one node is configured as the primary node, and the remaining nodes are configured as replica nodes. Redis is included with your HSTS installation.

In a HSTS high availability cluster:

  • Each node runs HSTS and uses the Redis database to manage security and file transfers. In a cluster configuration, all HSTS instances connect to the Redis database on the primary node, which is replicated to the replica nodes.
  • HSTS uses Hiredis, a Redis client, to interact with the Redis database.
  • Redis Sentinel monitors the status of Redis nodes, coordinates communication between nodes, and automatically promotes a replica node to primary if the current primary node fails.
HA Cluster

Requirements

To deploy a high availability HSTS cluster, your environment must meet the following requirements:

  • The cluster must consist of an odd number of nodes, with a minimum of three nodes.
  • A valid license or entitlement is required for each HSTS instance.
  • A load balancer or IBM Aspera Proxy is required to load balance file transfers across HSTS instances.

Features and limitations

Features

  • Redis database software that is included with your HSTS installation.
  • Redis persistence using point-in-time snapshots of the dataset.
  • Automatic failover from the primary node to a replica node if the primary node fails.
  • Eventual consistency across cluster nodes.

Limitations

  • High availability does not provide horizontal scaling or load-balanced throughput. To achieve these capabilities, combine HSTS HA with a load balancer or IBM Aspera Proxy.
  • Redis data is not sharded across nodes.
  • Bidirectional replication with conflict resolution is not supported.