Secondary Index Creation Time

secondary
index

#1

Are there any bench marks available regarding the secondary index creation timings.

I’m having a 6 node cluster with 100 Million Records with a replication factor of 2 (200 M records in total). The data is already there and we are using a hybrid configuration using a NVME SSD device, on i3.large instance in AWS. So any estimates on how long adding a secondary index of ‘String’ type, which will be present in all records be taking?

Also, during secondary index creation will other query or update operations be affected to the nodes in the cluster. If so what will be the impact? Will it affect operations based on primary key / other secondary key?

We are using Aerospike 3.13.0.8 Community Edition


#2

If you’re creating it in a live cluster, the cluster will remain online and regular requests will continue to work. The secondary index won’t be query-able until its done being built. How long it takes to built depends on your hardware and current load, number of records, the type/cardinality of the secondary index probably, as well as how many secondary index threads you assign. I don’t think we can give you a straight answer on that other than “test it out somewhere”.

See https://www.aerospike.com/docs/reference/configuration/#sindex-builder-threads - you can go as high as you want or as low as you want, just watch your resource utilization/latency.