Hi,
I have noticed that if the session is of the 'IPoE' type, then the CPU load is distributed relatively evenly.
irqtop - 2023-11-07 08:54:47 +0100 CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 cpuUtil: 37.3 0.2 2.0 7.9 16.9 26.6 total CPU utilization % %irq: 0.0 0.0 0.0 0.0 0.0 0.0 hardware IRQ CPU util% %sirq: 37.3 0.0 1.8 7.7 16.7 26.6 software IRQ CPU util% irqTotal: 59160 37 21402 26673 36761 49051 total hardware IRQs i 131: . . . . . 0 IR-PCI-MSI 376832-edge ahci[0000:00:17.0] i 138: 12870 . . . . . IR-PCI-MSI 524288-edge eth2-TxRx-0 i 139: . 13 . . . . IR-PCI-MSI 524289-edge eth2-TxRx-1 i 140: . . . . 4171 . IR-PCI-MSI 524290-edge eth2-TxRx-2 i 141: . . . . . 10870 IR-PCI-MSI 524291-edge eth2-TxRx-3 i 142: . . 4084 . . . IR-PCI-MSI 524292-edge eth2-TxRx-4 i 143: . . . 4246 . . IR-PCI-MSI 524293-edge eth2-TxRx-5 i 144: 0 . . . . . IR-PCI-MSI 524294-edge eth2 i 145: . . . . 24911 . IR-PCI-MSI 526336-edge eth3-TxRx-0 i 146: . . . . . 2407 IR-PCI-MSI 526337-edge eth3-TxRx-1 i 147: . . 9637 . . . IR-PCI-MSI 526338-edge eth3-TxRx-2 i 148: . . . 14548 . . IR-PCI-MSI 526339-edge eth3-TxRx-3 i 149: 16277 . . . . . IR-PCI-MSI 526340-edge eth3-TxRx-4 i 150: . . . . . 12825 IR-PCI-MSI 526341-edge eth3-TxRx-5 i 151: . 0 . . . . IR-PCI-MSI 526342-edge eth3 i NMI: 0 0 0 0 0 0 Non-maskable interrupts i LOC: 30011 21 7680 7878 7678 22947 Local timer interrupts i PMI: 0 0 0 0 0 0 Performance monitoring interrupts i RES: 0 1 0 0 0 0 Rescheduling interrupts i CAL: 0 0 0 0 0 0 Function call interrupts i TLB: 0 0 0 0 0 0 TLB shootdowns s TIMER: 56 17 19 33 52 68 s NET_TX: 24621 0 6984 7003 6625 19679 s NET_RX: 28047 13 13503 18542 28859 25139 s BLOCK: 0 0 0 0 0 0 s TASKLET: 0 0 0 0 0 0 s SCHED: 20 13 19 46 68 73 s RCU: 17 6 8 18 15 12
But when session is of the 'PPPoE' type, then:
- irqtop - 2023-11-07 09:03:07 +0100 CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 cpuUtil: 0.4 0.2 81.1 0.0 14.6 0.0 total CPU utilization % %irq: 0.0 0.0 0.0 0.0 0.0 0.0 hardware IRQ CPU util% %sirq: 0.0 0.0 81.1 0.0 14.4 0.0 software IRQ CPU util% irqTotal: 43 57 92309 64 25632 40 total hardware IRQs i 131: . . . . . 0 IR-PCI-MSI 376832-edge ahci[0000:00:17.0] i 138: 34 . . . . . IR-PCI-MSI 524288-edge eth2-TxRx-0 i 139: . 35 . . . . IR-PCI-MSI 524289-edge eth2-TxRx-1 i 140: . . . . 38 . IR-PCI-MSI 524290-edge eth2-TxRx-2 i 141: . . . . . 33 IR-PCI-MSI 524291-edge eth2-TxRx-3 i 142: . . 12613 . . . IR-PCI-MSI 524292-edge eth2-TxRx-4 i 143: . . . 36 . . IR-PCI-MSI 524293-edge eth2-TxRx-5 i 144: 0 . . . . . IR-PCI-MSI 524294-edge eth2 i 145: . . . . 25437 . IR-PCI-MSI 526336-edge eth3-TxRx-0 i 146: . . . . . 0 IR-PCI-MSI 526337-edge eth3-TxRx-1 i 147: . . 11434 . . . IR-PCI-MSI 526338-edge eth3-TxRx-2 i 148: . . . 0 . . IR-PCI-MSI 526339-edge eth3-TxRx-3 i 149: 0 . . . . . IR-PCI-MSI 526340-edge eth3-TxRx-4 i 150: . . . . . 0 IR-PCI-MSI 526341-edge eth3-TxRx-5 i 151: . 0 . . . . IR-PCI-MSI 526342-edge eth3 i NMI: 0 0 0 0 0 0 Non-maskable interrupts i LOC: 8 21 68261 28 154 5 Local timer interrupts i PMI: 0 0 0 0 0 0 Performance monitoring interrupts i RES: 0 0 0 0 1 0 Rescheduling interrupts i CAL: 0 0 0 0 0 0 Function call interrupts s TIMER: 2 10 136 16 44 4 s NET_TX: 0 0 43613 0 0 0 s NET_RX: 34 35 22461 36 49956 34 s BLOCK: 0 0 0 0 0 0 s TASKLET: 0 0 0 0 0 0 s SCHED: 3 12 83 18 58 3 s RCU: 4 9 16 9 13 2
What causes performance issues with a larger number of clients (one CPU becomes congested and network transfers significantly decrease).