From: Jeff <je...@pa...> - 2006-12-29 05:46:00
|
Thanks for the reply Gregg, Actually, that was going to be my next question. I have off-loaded all my ibutton reading to a seperate proxy machine. So the question is: Does the my main mh machine experience the delay? Or does it continue processing? I whole-heartly agree with avoiding the i/o delay, that's why I am asking. I assume the proxy machine takes the hit for delay, but the main machine does not? Is that right? I am still curious about the actual delay mh/proxy-mh takes to read the ibutton. If no one knows, I will figure it out in a couple days and report back. At 11:30 PM -0600 12/28/06, Gregg Liming wrote: >Hi Jeff, > >Quoting Jeff (12/28/06 11:03 PM): > >> All I really know is I have about 25 ibuttons and my mh systems seem >> to run really slow. Maybe it's my poor perl programming or maybe >> reading lots of ibuttons takes a long time? > >Relatively speaking, the 1-wire protocol is slow. This, like the cm11's >x10 protocol, is a reason that many people seek solutions that don't >ever require blocking the main mh instance's main loop. So, while >inefficient user-code programming can exacerbate the problem, the real >issue is relying upon slow synchronous protocols w/i mh's mail loop. >For some users, these issues are either minor or simply not noticed. I >personally perceive any delays to be unacceptable (because they never >scale) and instead rely on solutions that require zero I/O blocking (to >the mh main loop) and that never incur notable overhead. Using proxied >mh instances and/or "connected" daemons are broad examples of approaches >that decouple slow I/O. > >Regards, > >Gregg Jeff Pagel, MIEEE jeff AT pagel D0T net http://www.pagel.net (715) 359-8033 Cell:(715) 370-3476 or 370-disco ------------------------------------------------------------- |