Introduction to IBM Netezza Replication Services 2.0.0.0S

Netezza Performance Server (NPS) nodes are redundant, fault-tolerant systems. Adding IBM® Netezza® Replication Services for disaster recovery improves the fault tolerance by extending redundancy across local and wide area networks.

IBM Netezza Replication Services protects against data loss by synchronizing data on a primary system (the primary node) with data on one or more target nodes (replicas). These nodes make up a replication set.

When replication is active, Replication Services captures all SQL transactions that change data in the replicated databases on the primary NPS node and saves those transactions in a replication log. Examples of such transactions are those that insert, delete, or update data, that execute or update functions, or that add, delete, or update information about users. The replication log contents are automatically transmitted to all NPS replica nodes in the replication set. Each replica replays the transactions on its replicated databases, resulting in content that is the same as the primary's content.

Having a primary and one or more replicas can provide data availability even if a NPS node is offline or unreachable. The replica nodes act as “warm standbys” for system failure or planned downtime at the primary. If data on a NPS node is lost, such as due to a network outage or multiple hardware failures, the replication set has sufficient redundant copies to recover (you need at least one healthy copy). Because each NPS node in the replication set stores a copy of the data and the replication log, if a node is damaged and its local data is lost, the latest data is either in the replication log or on one or more of the Netezza NPS nodes. Therefore, if one node fails, processing of replicated data can continue at another node with a bounded loss of data.

The replication approach to disaster recovery has an additional benefit. Because multiple copies of the database are distributed across several Netezza NPS nodes, the configuration supports more SQL queries running against the same data. Replica nodes also provide concurrent BI and reporting access to synchronized data stores, thus allowing workload distribution, balancing, and optimization.

Because no transformation or other CPU-intensive operations (for example, log processing, intrasite data transfers, or TCP/IP network negotiation) are performed on the Netezza NPS node, replication resource usage is very low.

The following figure shows an implementation of IBM Netezza Replication Services.
Figure 1. Replication environment
In the sample implementation that is shown in the diagram. there is a primary node, typically in a local data center. All writes (loads, SQL, and data from analysis and reporting applications) are sent to the NPS instance in the primary node. Loads come from files and an ETL system. SQL consists of sporadic global updates. The NPS instance in the primary node communicates with a replication RQM through a LAN. All writes from the replication queue manager of the primary node flow to the replication queue manager of the replica node through a WAN. The replication queue manager of the replica node is connected to the NPS instance of the replica node through a LAN. The replica can be on a local or remote site.