Using average replication latency to influence routing query transactions

1. A method of routing workload data to a plurality of processing sites, comprising:

defining, using a processor, a maximum replication latency threshold and a reset threshold;

providing an update workload to a first of the plurality of processing sites;

providing the update workload from the first of the plurality of processing sites to another of the plurality of processing sites;

estimating a replication latency of the workload for the plurality of sites using a difference in time between completion of a transaction on the update workload at a first of the plurality of processing sites and a completion of a replication process of the transaction at the other of the plurality of processing sites;

switching to an acute replication latency state for the plurality of processing sites when the estimated replication latency of the workload rises above the defined maximum replication latency threshold, wherein a query workload is provided to only the first of the processing sites in the acute replication latency state; and

switching out of the acute replication latency state when the estimated replication latency falls below the defined reset threshold.

2. The method of claim 1 , wherein the reset threshold is less than the maximum replication latency threshold.

3. The method of claim 1 , wherein switching out of the acute replication latency state further comprises switching to a normal replication latency state of the plurality of sites.

4. The method of claim 3 , wherein the query workload is distributed to each of the processing sites in the normal replication latency state.

5. The method of claim 3 , further comprising defining at least one other replication latency state over a range of replication latency values between the normal replication latency state and the acute replication latency state, wherein the at least one other replication latency state favors use of a single-site distribution configuration.

6. The method of claim 5 , further comprising switching from the normal replication latency state to the at least one other replication latency state when the estimated replication latency lies within the selected range of replication latency values defining the at least one other replication latency state.

7. The method of claim 1 , further comprising estimating the replication latency by obtaining an average of replication latencies for the plurality of processing sites over a selected time interval.

8. A data processing system, comprising:

a first processing site configured to perform a transaction on an update workload;

a second processing site configured to perform a replication process of the transaction at the first processing site;

a workload distributor for distributing a query workload to one or more of the first processing site and the processing second site;

a transaction replay program configured to send a copy of update workload from the first site to the second site; and

a processor configured to:

define a maximum replication latency threshold and a reset threshold for the system,

estimate a replication latency of a workload in the system from a difference in time between completion of the transaction at the first processing site and a completion of the replication process at the second processing site, and

switch the system to an acute replication latency state when the estimated replication latency rises above the maximum replication latency threshold and switch out of the acute replication latency state when the estimated replication latency falls below the reset threshold, wherein the query workload is distributed to only the first processing site when the system is in the acute replication latency state.

9. The system of claim 8 , wherein the reset threshold is less than the maximum replication latency threshold.

10. The system of claim 8 , wherein the processor is further configured to switch out of the acute replication latency state directly into a normal replication latency state of the plurality of sites when the estimated replication latency falls below the reset threshold.

11. The system of claim 8 , wherein query workload is distributed to the first processing site and the second processing site when the system is in the normal replication latency state.

12. The system of claim 10 , wherein the processor is further configured to define at least one other replication latency state over a range of replication latency values between the normal replication latency state and the acute replication latency state, wherein the at least one other replication latency state favors use of a single-site distribution configuration.

13. The system of claim 12 , wherein the processor is further configured to switch from the normal replication latency state to the at least one other replication latency state when the estimated replication latency lies within the defined range.

14. The system of claim 13 , wherein the processor is further configured to estimate the replication latency as an average of replication latencies of the workload over a selected time interval.

15. A method of operating a system of independent processing sites, comprising:

defining, using a processor, a maximum replication latency threshold and a reset threshold of the system;

providing an update workload to a first of the plurality of processing sites;

providing the update workload from the first of the plurality of processing sites to another of the plurality of processing sites;

estimating a replication latency of a workload in the system as a difference in time between completion of a transaction at a first of the processing sites and a completion of a replication process of the transaction at another of the plurality of sites;

switching the system to a first replication latency state when the estimated replication latency rises above the maximum replication latency threshold, wherein a query workload is provided to only the first of the processing sites in the acute replication latency state; and

switching the system out of the first replication latency state to a second replication latency state when the estimated replication latency falls below the reset threshold.

16. The method of claim 15 , wherein the reset threshold is less than the maximum replication latency threshold.

17. The method of claim 15 , wherein the first replication latency state includes an acute replication latency state and the second replication latency state includes a normal replication latency state.

18. The method of claim 17 , further comprising defining at least one replication latency state over a selected range of replication latency values between the normal replication latency state and the acute replication latency state.

19. The method of claim 18 , further comprising switching from the normal replication latency state to the at least one other replication latency state when the estimated replication latency lies within the selected range of replication latency values.

20. The method of claim 15 , wherein estimating the replication latency of the workload further comprising obtaining an average of replication latencies for the workload over a selected time interval.