High Availability and IMCS

This chapter describes high availability features supported with IMCS.

IMCS in TAC

When using IMCS in Tibero Active Cluster (TAC), every running instance has its own IMCS area. By default, objects populated in a TAC environment are distributed across all nodes in the cluster.

In Tibero, the INMEMORY_SIZE parameter for every TAC node must be set to an equal value.

In-Memory Distribution

You can use DISTRIBUTE to specify how data of an In-Memory object is distributed to each TAC node. The following describes each DISTRIBUTE option.

DISTRIBUTE option

Option	Description
AUTO	Distributes data equally to each node in accordance with Tibero internal rules. (Default)
By Rowid Range	Distributes data to each node by rowid range. Currently not supported in Tibero.
By Partition	Distributes data to each node by partition. Currently not supported in Tibero..
By Subpartition	Distributes data to each node by subpartition. Currently not supported in Tibero.

In-Memory Duplication

You can use DUPLICATE for an In-Memory object to specify whether and how each TAC node stores the copy of data.

DUPLICATE option

Option	Description
NO DUPLICATE	No data copy of In-Memory data. (Default)
DUPLICATE	One copy for each In-Memory object data. When populating an In-Memory object specified with the DUPLICATE option, the same data is stored in 2 TAC nodes. Currently not supported in Tibero.
DUPLICATE ALL	As many copies for each In-Memory data as the number of nodes. Every instance stores the copies of an In-Memory object in its own IMCS area.

Parallelism in TAC

Since In-Memory data populated in a TAC environment is distributed across all TAC nodes by default, queries for In-Memory data are executed in parallel. For example, when you perform an In-Memory scan for a table populated in a 3-node TAC environment, the database reads data from every corresponding table stored in each TAC. This means that In-Memory scans in a TAC environment are performed in parallel, with the degree of parallelism (DOP) determined based on the number of nodes.