Summary: This page provides a brief description for the different replication topologies
Primary Backup TopologyA Primary-Backup topology (also known as Active-Passive) consists of a group of space instances that are part of the same replication group, one of the space instances is elected and serves as the primary space instance of the group. The application interaction is done with the primary instance while the other instances are serving as the backups of the primary. The backup space instances are an exact replica of the primary and if the primary instance fails they are in place to replace it when such failure occurs without losing any data and in a transparent manner to the proxy connected to the spaces. In order to maintain this reliability with no data loss, the default replication mode between the space instances in this key is Synchronous Replication. The common usage of this topology is to have it together with a partitioned space, such that each partition is isolated in terms of replication. This means that each partition consists a separate replication group of space instances such that in each partition only one primary will be elected and all the other members of the partition will serve as its backups. There is no replication taking place across partitions. After a primary is elected, each remaining backup member will first recover its state from the primary before becoming available.
Primary Backup with a Mirror TopologyThis topology extends the previous one by having a mirror service that replicates all the modifying operations (Write/Update/Take in and out of transaction) to an external data source. The most common external data source is a database that is used for storing all some or all of the data in the cluster.
Active-Active TopologyThis topology consists of a group of space instances that are part of the same replication group. Unlike primary backup, all of the space instances are active and the application interacts with all of them. The common use case for this topology is to share reference data across instances. Each space instance replicates its data to all the other members, thus making sure that all the space instances will contain the same data set. This topology is more commonly used with Asynchronous Replication. Since each space instance is active, it is up to the application to make sure that there is no conflicting operations, by not updating/changing the same object in two different space instances at the same time. With this topology, each new space instance that is started, first tries to find an already existing active member in the replication group, and if such a member is found it will first perform recovery from that member before becoming active and available.
Load BalancingSince all the space instances hold the same data, there are multiple load balancing strategies the space proxy can choose from. By default a sticky strategy is used, i.e. all operations are performed on the same space instance. For more information on load balance strategy selection and configuration, see Proxy Connectivity. |
![]() |
GigaSpaces.com - Legal Notice - 3rd Party Licenses - Site Map - API Docs - Forum - Downloads - Blog - White Papers - Contact Tech Writing - Gen. by Atlassian Confluence |