[e2e] TCP and UDP Processing Overhead Delays.

[Craig Partridge]

Written concurrently with their paper is:

    Partridge 7 Pink, A Faster UDP, IEEE/ACM Trans. on Networking,
    August 1993.

It discusses a number of ways to make a UDP implementation faster.

Craig

In message <5BBECACEE2CBD94BA0C75229E9E81D6B02A54D02 at xch-sw-01.ds.boeing.com>, 
"EXT-Edwards, William H" writes:

>Does anyone have any or know of any current figures that can update the 1993
>work of Kay and Pasquale in analyzing TCP and UDP Overheads to reduce or
>preclude data touching?  Their work using DECstation 5000/200 workstations
>with the IO bus monitored by a HP-1625B with 40 nanosecond resolution to
>check and compare the overhead processing time delays and the amount of data
>touching for Checksums, DataMove (user-to-kernel, device copy, and cache
>clear), DataStruct (socket buffers, IP Defragmentation queue, device
>interface queue), ErrorCheck, Mbuf, OpSys Overhead (socket support and
>synchronization), ProtSpec, and Other (total processing time required
>compared to time required in the other previously mentioned areas).  
>
>William H. Edwards
>Systems Engineer
>Booz-Allen & Hamilton
>(703) 872-4289 (Voice)
>(703) 872-4254 (Fax)
> 
>