6 Send-side processing
Until now, we have explored affinity-based scheduling of receiveside
UDPIIPIFDDI processing. How does the scheduling technique
apply to send-side processing, and what is the comparative
performance? To explore this, we revisited the original, unparallelized
z-kernel implementation, and examined the send-side
execution path from the top of the stack down through the UDP,
IP, VNET and FDDI protocols and into the simulated FDDI device
driver. The goal was to identify the protocol and stream-specific
data structures written during a send operation, as a first step toward
developing send-side PLP and CLP implementations.