Looking for an alternative for matlab's fread() - multiple computer communication

I have a system of 2 computer that run seperate matlab session. In principle, I want to send a 'trigger' (6 byte double packet) to one computer just after the other computer finishes to run some code. I wish to get the trigger in fine time resolution (<100ms) and to be able to measure the latency if needed. Up to know, I used UDP communication with the following scheme:

computer 1 - ipA:
      % sending the trigger
      data = clock;
      fwrite(ipB,data,'double');

meanwhile, the other computer scans for an non-empty packet:

computer 2 - ipB:
      % create timer object for triggering       
      trigger = InitializeMyTimer(ipA);
      triggertime= trigger.UserData;

when the timer has the following structure:

function [ triggObj ] = InitializeMyTimer(ipData)
% InitializeMyTimer() is a function to manage the reception of the a
% trigger from computer ipA
  triggObj = timer;
  triggObj.Period = 0.01; % 1 ms time between executions of the TimerFcn 
  triggObj.ExecutionMode = 'fixedRate';
  triggObj.TimerFcn       = @mytimer_cb;
  triggObj.BusyMode       = 'drop';
  triggObj.TasksToExecute = 1e3; % this determines the timeout of the run in this gui
  triggObj.UserData       = ipData;
  start(triggObj);
  wait(triggObj);
end

function mytimer_cb(obj, event)
% timerFunc stops running when we recieve a packet of data
  clocktime= fread(obj.Userdata ,6,'double');
    if ~isempty(clocktime)
      obj.Userdata = clocktime;
      stop(obj);
  end;  
end

'fread' seems to be the bottleneck of this function since it takes at least 1 sec to run

clocktime= fread(obj.Userdata ,6,'double');

Further, I don't really found a good way to measure the send/receive latency. I though about several directions -

(1) send an analogoutput to a NI board connect to the ipB and read it from MATLAB

(2) try somehow to run multiple executions of 'fread' simultaneously.

(3) use serial communication between computer with a RS232 null mode cable and to read the trigger this way.

What should be the most promising technique? What do you guys suggest as an alternative?

Thanks, Noam