Inefficiency in preallocation algorithm for flq while using multiple subnets
While using flq
allocator Kea is able to preallocate 18M leases in just 12seconds (using our setup in the lab), which is very good result. But while Kea is configured with multiple subnets using flq this preallocation is made one by one without multithreading.
preallocation for 1 subnet with /8 pool: preallocation for 8 subnets with /8 each:
As you can see in second chart 36-B preallocation for 8 subnets takes 120 seconds and uses just one CPU (100% CPU usage). While Kea is configured to use 8 threads. This could be done in parallel to reduce overall startup time.