Variable configurations for workload distribution across multiple sites

摘要

Embodiments include a method, system, and computer program product for maintaining continuous availability. An embodiment of the present invention includes receiving a unit of work data. The unit of work data is distributed among the plurality of workloads using either an active/standby configuration, an active/query configuration, an active/partitioned configuration, or an active/active configuration. A primary workload is selected from a plurality of workloads to process the unit of work data based on a workload distribution rule. It is determined if the primary workload is capable of processing the unit of work data based on user configurable settings. If the primary site is capable of processing the unit of work data then the unit of work data is transmitted to the primary workload. Otherwise the unit of work data is transmitted to a secondary workload. The unit of work data is replicated between the primary workload and the secondary workload.

主权项

1. A method for maintaining continuous availability, the method comprising:
establishing a plurality of workload distribution rules, comprising a workload distribution rule for each of a plurality of workloads, wherein a first workload distribution rule applies to a first workload and a second workload distribution rule applies to a second workload; receiving four or more units of work data, comprising a first unit of work data for a first workload replicated in a plurality of first replicas across a plurality of sites, and a second unit of work data for a second workload replicated in a plurality of second replicas across the plurality of sites; selecting a primary replica of the first workload from among the plurality of first replicas to process the first unit of work data based on the first workload distribution rule; selecting a primary replica of the second workload from among the plurality of second replicas to process the second unit of work data based on the second workload distribution rule; wherein the primary replica of the first workload and the primary replica of the second workload exist simultaneously and are located on distinct sites from each other among the plurality of sites; determining, by a computer processor, based on one or more user configurable settings, that the primary replica of the first workload is capable of processing a third unit of work data for the first workload, the third unit of work data being among the four or more units of work data; transmitting the third unit of work data to the primary replica of the first workload responsive to determining that the primary replica of the first workload is capable of processing the third unit of work data; and replicating the third unit of work data between the primary replica of the first workload and a secondary replica of the first workload, the secondary replica being from among the plurality of first replicas; wherein each of the four or more units of work data is distributed among the plurality of sites according to the first workload distribution rule using one of:
an active/standby configuration,an active/query configuration,an active/partitioned configuration, andan active/active configuration; wherein a fourth unit of work data of the four or more units of work data is distributed among the plurality of first replicas across the plurality of sites using the active/query configuration, and wherein the method further comprises:
inspecting the fourth unit of work data;determining whether the fourth unit of work data is a query transaction;transmitting the fourth unit of work data to the primary replica of the first workload if the fourth unit of work data is not a query transaction; andif the fourth unit of work data is a query transaction:
determining replication latency for the fourth unit of work data from the primary replica of the first workload to a third workload designated for queries;if the replication latency exceeds a latency policy threshold, routing the query transaction to the primary replica of the first workload; andif the replication latency does not exceed the latency policy threshold, routing the fourth unit of work data to the third workload designated for queries.