Hi, we have an Aerospike cluster in production with dozens of nodes. We would like to know if there is a way that our application can have the knowledge that which nodes the partition belongs to. We can group requests with the same nodes using this information to improve performance. Thanks!
The methods that determine which node the partition belongs to are located in Partition.java.
A similar question was also asked here where we provided example code.
By the way, how do you expect to increase performance with this information?
The client already offers batch APIs which will further subdivide into batches to particular nodes holding the data.
Hi kporter, we are now making batch-requests with small batch sizes such as 10~50 due to some limitations. We think if all requests in one batch were sent to only one physical node, the P99 latency would be better. For example, in the worst case, the 10 requests in one batch will be subdivided into 10 sub-batches and send to 10 physical nodes.
This topic was automatically closed 6 days after the last reply. New replies are no longer allowed.