Re: TCP speed

[csapuntz@cisco.com]

"Chee-Siong, Lee" <cslee@FalconStor.com.tw> writes:

>     Sorry for poor description..
> 
>    A multithread program will use different cpu in different thread. For
> example, we use one
> cpu for receving packets to receive queue and the other CPU for sending
> packets from sending
> queue. Hence we can have more CPU power in iSCSI path. An UDP program with
> its sending
> and receiving can archive this situation. I think this will give a
> significant performance than TCP if
> TCP can't use both CPUs proccesing power.

UDP is missing the following featuers:

1) FIFO ordering of SCSI commands across the wire
2) Congestion control
3) Segmentation of payloads (unless you want to set off an
IP fragmentation grenade)
4) Discovery when/recovery from dropped segments
5) Protection against lingering packets in the network

Unfortunately, by the time you put in the features, you have added
dependencies between the transmit and receive paths (mostly related
to acknowledgements) which I believe are on the order of TCP in
complexity.

-Costa