Linux Loopback Performance with Tcp_Nodelay Enabled

TCP loopback connection vs Unix Domain Socket performance

Yes, local interprocess communication by unix domain sockets should be faster than communication by loopback localhost connections because you have less TCP overhead, see here.

Set TCP_QUICKACK and TCP_NODELAY

There's no direct relationship between those two options, they are just for different purposes.

TCP_NODELAY is intended to disable/enable segment buffering so data can be sent out to peer as quickly as possible, so this is typically used to improve network utilisation. TCP_QUICKACK is used to send out acknowledgements as early as possible than delayed under some protocol level exchanging, and it's not stable/permanent, subsequent TCP transactions (which may happen under the hood) can disregard this option depending on actual protocol level processing or any actual disagreements between user setting and stack behaviour.

NOTE TCP_NODELAY is portable while TCP_QUICKACK is not (only works under Linux 2.4.4+).

in linux, does routing take different path on loopback vs IP assigned to NIC

The hardware does not get involved.

As soon as the routing function knows that the destination address is local, the packet is switched to ingress path. Which is incidentally why sniffers can't capture such packets, because that hook happens to be after the point of this decision.

I need a TCP option (ioctl) to send data immediately

IIUC, setting the TCP_NODELAY option should flush all packets (i.e. tcp.c implements setting of NODELAY with a call to tcp_push_pending_frames). So if you set the socket option after each send call, you should get what you want.

---

Related Topics