Instead of allocating new vectors for each partition, simply partion the nodes in place and pass the corresponding ranges to the next phase.
Instead of allocating new vectors for each partition, simply partion the nodes in place and pass the corresponding ranges to the next phase.