Performance
Showing results for 
Search instead for 
Do you mean 

MPE/iX 7.5 PINs 4-9 proc(stf_port_facility_process)

SOLVED
Go to Solution
Trusted Contributor Trusted Contributor

MPE/iX 7.5 PINs 4-9 proc(stf_port_facility_process)

Glance on 7.5 system shows pins 4 to 9 in the A queue with priority 0. The Glance Process screen identifies them as "proc(stf_port_facility_process)". The CPU % of each can go as high as 7% when the system (N4000 4 way 750) is busy. This seems a significant overhead, what is the purpose of these processes?

P4 SYS MANAGER.SYS 4 A 0 1.3% 0.0 0 0.0 MSG
P5 SYS MANAGER.SYS 5 A 0 1.3% 0.0 0 0.0 MSG
P6 SYS MANAGER.SYS 6 A 0 1.3% 0.0 0 0.0 MSG
P7 SYS MANAGER.SYS 7 A 0 1.3% 0.0 0 0.0 MSG
P8 SYS MANAGER.SYS 8 A 0 1.3% 0.0 0 0.0 MSG
P9 SYS MANAGER.SYS 9 A 0 1.3% 0.0 0 0.0 MSG
1 ACCEPTED SOLUTION
Highlighted
Honored Contributor Honored Contributor

Re: MPE/iX 7.5 PINs 4-9 proc(stf_port_facility_process)

Hello Fred,

These pins #4 to #9 are processes for the native Fibre Channel IO. Unlike previous IO drivers, the FC 'driver' cannot run on the ICS (Interrupt Control Stack) so these extra processes were added to handle some central aspects of FC IO (the Interprocess Communication parts).

If the system has a lot of disks on FC then it would not be a surprise there are consuming some amounts of CPUs. For ICS drivers, the CPU for IO was calculated and assigned to the process requesting/receiving the IO, but now these 6 processes get them all.
King is the customer!
1 REPLY
Highlighted
Honored Contributor Honored Contributor

Re: MPE/iX 7.5 PINs 4-9 proc(stf_port_facility_process)

Hello Fred,

These pins #4 to #9 are processes for the native Fibre Channel IO. Unlike previous IO drivers, the FC 'driver' cannot run on the ICS (Interrupt Control Stack) so these extra processes were added to handle some central aspects of FC IO (the Interprocess Communication parts).

If the system has a lot of disks on FC then it would not be a surprise there are consuming some amounts of CPUs. For ICS drivers, the CPU for IO was calculated and assigned to the process requesting/receiving the IO, but now these 6 processes get them all.
King is the customer!