Generally speaking, the more requests that are processed in parallel, the higher the number of requests that are processed over all. Given the appropriate amount of hardware, processing N requests concurrently is typically faster than processing N requests serially.
In PingFederate, there are two main pools of threads that control the level of concurrent user requests: Acceptor Threads and Server Threads. Acceptor threads receive the HTTPS requests, and pass those requests on to available server threads to be processed.
It is important to note that this topic serves as a guideline for optimizing the concurrency of your deployment. On a large system with multiple CPU (or cores), a thread pool that is too small will under-utilize the available processor resources. A thread pool that is too large can cause the system to become flooded and unusable.
A good target for the CPU is between 60%-80% utilization when under nominal (reasonably expected) user load. This way, CPU resources are not under-utilized while still allowing room for occasional load spikes. The level of concurrency in PingFederate may need to be decreased (or even increased) depending on the system's configuration (the adapters in use), available memory, and other processes competing for resources. All deployments are different. This topic therefore serves as a guideline of where to start when tuning the server.