You can use the Bulk API to process jobs either in serial mode or in parallel mode. Processing batches serially means running them one after another, and processing batches in parallel means running multiple batches at the same time. When you run a Bulk API job, processing more of its batches in parallel means giving that job a higher degree of parallelism, which in turn gives your overall run better data throughput.
When you boost your degrees of parallelism by optimizing your Salesforce organization's architecture and loading your large volumes of data properly, you can end up with data loads and integrations that perform orders of magnitude faster than they otherwise would.
From Maximizing Parallelism