Re: Dedicated Z-wave sites?

["Robert Green" <ROBERT_GREEN1963@xxxxxxxxx>]

"Marc_F_Hult" <MFHult@xxxxxxxxxxxxxxxxxxxxx> wrote in message
news:s4d3o2d2crd3q055rtea90nbm3spkaf1fd@xxxxxxxxxx
> On Thu, 14 Dec 2006 15:13:05 -0500, "Robert Green"
<ROBERT_GREEN1963@xxxxxxxxx>
> wrote in message  <cJSdnUvB1ri3LBzYnZ2dnUVZ_u63nZ2d@xxxxxxx>:
>
> >"Dean Roddey" <droddey@xxxxxxxxxxxxxxxx> wrote in message
news:jlhgh.24869
> >
> ><stuff snipped>
> >
> >> I dunno. I think that you are maybe stressing out over something that's
> >> not nearly as big an issue as you are thinking. Yes, there are crappy
devcie
> >> drivers out there. And if you configure a machine with a fairly random
set
> >> of hardware, you can have problems. OTOH, there are quality products
out
> >> there that work, and they become known well enough. A machine that is
set
> >> up with good quality hardware and drivers, which isn't used as a daily
use
> >> machine (i.e. it's configuration is not changed and web surfing isn't
done
> >> on it and things that aren't needed are turned off in the OS, i.e. a
> >> standard kiosk style touch screen client or a server in the closet) can
be
> >> stable for years without problems.
> >
> >That's a lotta futzing and "kid gloving."  It likely means no one will be
> >plugging IP cams and Ethernet switches into that same network which could
> >limit usefulness in a big way, at least in terms of HA.
>
> Bobby,
>
> I don't understand why you would say this. One could kill a network in a
variety
> of less esoteric ways -- for instance  shorting two wires in the CAT5.
What does
> this have to do with the stability of an HA server? If (eg) an IP camera
breaks
> a newtork, the problem is with the camera.

It's got to do with why HA programs like CQ are slow to gain acceptance.  I
just bought a Panasonic Netcam from Smarthome.  I want to plug it into my
system and email pictures to my cellphone.  Wouldn't the natural place to
install such a beast be the HA server?  If a net cam has a *really* bad
driver, it's going to blow up CQ.  For CQ to be attractive to me, it would
have to implement some of the features that I referred to in the NOOKS
citation.  Many people cite the decline of Homeseer reliability as
coinciding with the heavy reliance on "plug ins" - really another form of
device drivers.  Something I saw today said that 6/7ths of the existing
Linux codebase now consists of device driver code.

If we are to follow best practices, it now sounds as if we're talking a
client to go along with that server.  The costs of the dedicated HW alone
are now starting to exit the home market price range.  But Dean says he's
not targeting that market, so maybe it's no problem for anyone needing CQ's
features.  If X-10 has taught me anything, it's that price matters.  A lot.

> It is conventional IP Best Practice to allocate servers on an at-least-
> one-per-function basis.

Business practice, sure.  Home network practice, not so sure.

Server farms have a very low SAF. :-)  So, IMHO, does more than one PC per
room.  Compromises are far more likely to occur in home installations where
people aren't likely to be IT pros.

> Why should HA servers deviate from what are established
> best practices? An HA server is a discrete function. It is sound practice
to
> allocate at least one CPU to it --  not "kid gloving".

I dunno.  I have plenty of machines that serve more than dual purposes.
Were I to strictly devote one PC per application, I would be running a
server farm.  While it may certainly be best practices for a business, size
and money constraints often dictate that more than one application is going
to live on a server, especially on a home network.  It's been my contention
that while it should be possible, it's usually only achievable by a PC guru
that can configure a server the way Perlman can strum a violin. Good
software is able to interact with other good software without bringing down
the house.  The more a program has to rely on isolation to function
properly, the less appeal it will have to people.

> One can over-tax any machine ever built, whether by throughput, CPU cycles
or
> complexity of simultaneously running software. I can say from practical
> experience with several different mini-ITX machines (which you have
expressed an
> affinity for)  that they can run out of CPU cycles with only a few HA
tasks --
> trying to also simultaneously run general purpose computing or networking
chores
> can bring them to their knees long before software interactions become a
major
> problem.

"Can bring" is correct but should such chores choke a 1GHz CPU or is the
software bloated, inefficient and buggy?  Dean says, and rightly so, that
you can find the perfect mix of peripherals, PC, apps and OS.  The questions
I have are who's finding it, how do they recognize unless they are PC
experts and what is the bottom line cost?

One certainly can overtax any machine ever built but the high end standard
today is 2 cores and 2GB of memory at 2GHz clock speeds.  That should be
able to handle some pretty serious applications - concurrently.  From what
I've been reading, 4, 8 and 16 multicore chips are in the pipeline although
the MS OS's are woefully ill-equipped to handle true parallel
multiprocessing.  But that's beside the point.

My concern is that an HA server that has to be sequestered to operate
properly is by the very nature of isolation, going to be that much hard to
connect to plug in Ethernet appliances, IP cameras, daily on-line weather,
traffic and email data and all the other devices that are in the pipeline
now.   Lots and lots of "appliances" are coming with RJ-45 jacks and
embedded servers to allow PC configuration.  If I'm running an HA server and
new HA products come on line, it seems the HA server is the place they
should go.  If there server is only going to stay stable through careful
selection of HW and SW additions, how does the end user evaluate the quality
of their drivers?  What are the security implications?  This all rapidly
gets so esoteric that all but the faithful roll their eyes and sit back and
wait for something simpler and cheaper from Microsoft.

Let me ask you this.  Best IP practices dictate grandfather, father, son
rotating backups, some of which should be stored off site, all of which
should be tested to insure accurate data restoration.  How many businesses
do you think do it?  How many home network administrators?  There's the
ideal and theoretical world and then there's the jungle of the real world.

--
Bobby G.