TBB parallel pipeline on mix of IO and compute workload

Given million query vector, for each query I want to find similar vectors among some DB of vectors and I want to parallelize this workload using TBB. The query consists of transforming the candidate vectors from DB using several functions. I wrap each function as separate tasks and use pipeline_parallel with filters as tasks. I want to ask whether parallel_pipeline would be actually good for this kind of workload where I have the mix of IO and compute.