Revision Two: Redesigning for Fewer Connects

In this revision, the sample application is redesigned to eliminate unnecessary connects.


This examples of the application also provide intentionally poor performance, in order to illustrate performance improvements possible with changes to code. Do not use this code sample in your application; it is for illustration purposes only.


ComputeNext( Map );
bind( s, ... );
connect( s, ... );
for(int i=0 ; i < ROWS ; ++i)
  for(int j=0 ; j < COLS ; ++j)
    BYTE tmp[3];
    tmp[0] = i;
    tmp[1] = j;
    tmp[2] = Map[i][j];
    send( s, tmp, 3 );
    recv( s, &byRet, 1 );
closesocket( s );

Remaining Problems

The changes in Revision Two redesigned the application to make only one connect per update. The application still includes the following performance issues:

  • The application is still serialized and chatty.
  • The application still has a fat design; there are many sends that require no operation in this design.
  • The sends are still only 3 bytes, which is poor streaming.

Key Performance Metrics

The following performance metrics are expressed in Round Trip Time (RTT), Goodput, and Protocol Overhead. See the Network Terminology topic for an explanation of these terms.

This version reflects the following performance metrics:

  • Cell Time — 1*RTT
  • Goodput — 4 bytes/RTT
  • Protocol Overhead — 96.8%

Improving a Slow Application

Network Terminology

The Baseline Version: A Very Poor Performing Application

Revision 1: Cleaning up the Obvious

Revision 3: Compressed Block Send

Future Improvements