UDB-TX has a limit on the maximum number of concurrent connections, and when this number is reached, new connections are rejected. However, increasing this maximum number of connections increases resource consumption and negatively affects overall system performance.UDB-CLUSTER also has a limit on the maximum number of connections, but additional connections are queued instead of returning an immediate error. Of course, it can be configured to return an error when the connection limit is exceeded.

When the primary library in the cluster is unavailable, UDB-CLUSTER can detect and activate the backup library to realize failover, thus ensuring business continuity and realizing high availability of production.

UDB-CLUSTER can perform online recovery of database nodes by executing a single command. When online recovery is used in conjunction with automatic failover, nodes separated by failover can be automatically attached as standby nodes. New UDB-TX nodes can also be synchronized and attached.

Querying the cache in memory allows a pair of SELECT statements and their results to be saved.If the same SELECT enters, UDB-CLUSTER returns the value from the cache.Using the in-memory cache is very fast since it does not involve SQL parsing and access to UDB-TX.On the other hand, in some cases it may be slower than the normal path because it adds some overhead to storing cached data.

Watchdog can coordinate multiple UDB-CLUSTERs to create a robust clustered system that avoids single points of failure or brain fission. To avoid brain cracking, at least 3 UDB-CLUSTER nodes are required.Watchdog can perform life checks against other UDB-CLUSTER nodes to detect UDB-CLUSTER failures.If the active UDB-CLUSTER goes down, the standby UDB-CLUSTER can be upgraded to active and take over the Virtual IP.

UDB-CLUSTER provides a unified back-end database UDB-TX connection pooling function to reduce the overhead caused by the establishment of connections, while increasing the throughput of the system.

When the database UDB-TX is running in replicated or master/standby mode, the SELECT statement running on any node in the cluster can return consistent results. UDB-CLUSTER can distribute the query statement to each database in the cluster, thus increasing the throughput of the system. Load balancing is applicable to read-only scenarios.