Fortress

[hucker]

Is it really virtualization or just putting everything on one machine? I have a dual core, RAID based (4 drives) server machine that runs all the music, the home automation, wife's PC and laptop backups, and connection to a bunch of serial devices. This machine draws 80W. The UPS, doing nothing draws 20W! I have two NAS drives that idle along at 5W each and draw 20W when backing up for about 5-20minutes each day. My total server consumption probably around 115W, with the server 95% of the time doing nothing. I measured this with my Killawatt meter.

It seems in a home automation/server scenario the power benefits of virtualization aren't so great unless I'm missing something (which at my advanced age is likely!).

I *think* on your ROI calculation you are assuming that your power draw equals your power supply size. That is not the case, at least for me. My machine has a 450W supply but only draws 80W in the idle state (at boot time I see it go to about 130W). I suspect that your 650W PS is similar. I could be wrong here if you have some loading that I don't have such as streaming DVD's or capturing surveillance video streams. If you put a killawatt meter and watch for a while I think you will find that you are drawing MUCH less than you think say by a factor of 5 which could make your payback something around 9years. Again YMMV, I'd love to see the measured numbers so we can keep the discussion as accurate as possible.

 

[kwilcox]

Is it really virtualization or just putting everything on one machine? I have a dual core, RAID based (4 drives) server machine that runs all the music, the home automation, wife's PC and laptop backups, and connection to a bunch of serial devices. This machine draws 80W. The UPS, doing nothing draws 20W! I have two NAS drives that idle along at 5W each and draw 20W when backing up for about 5-20minutes each day. My total server consumption probably around 115W, with the server 95% of the time doing nothing. I measured this with my Killawatt meter.

It seems in a home automation/server scenario the power benefits of virtualization aren't so great unless I'm missing something (which at my advanced age is likely!).

I *think* on your ROI calculation you are assuming that your power draw equals your power supply size. That is not the case, at least for me. My machine has a 450W supply but only draws 80W in the idle state (at boot time I see it go to about 130W). I suspect that your 650W PS is similar. I could be wrong here if you have some loading that I don't have such as streaming DVD's or capturing surveillance video streams. If you put a killawatt meter and watch for a while I think you will find that you are drawing MUCH less than you think say by a factor of 5 which could make your payback something around 9years. Again YMMV, I'd love to see the measured numbers so we can keep the discussion as accurate as possible.

The 4 machines I virtualized were:

My HCA server (XP)
My Active directory domain controller (W2K3 Server R2)
My Linux based PBX (CentOS 5)
My media server (XP)

We tried combining the iTunes server with HCA with but ran into performance issues. Each system was also optimized for the type of OS it ran on so crossing with others was difficult. The original server (a dual Pentuim II Dell Poweredge 1300) ran pretty hot because the PII chips were very power hungry and I kept the mirrored scsi disks spinning all the time to increase longevity. By comparison, the T605 uses the latest AMD processors so probably isn't using all of the 650W powersupply. However, the $60 bump on my electric bill makes me think that it is. Hard to tell; I'll put a watt meter on the old machines that are now de-commissioned to see what they draw. It'll be hard to test the T605 without using a passive meter since it takes a while to power down/back up.

My point was simply this (as stated above). If you have multiple PCs running at your house then this becomes something to look at.

 

[dBeau]

Hucker,

I'd have to agree with your numbers. What I think of as a typical desktop class machine (one maybe two drive, and mediocre mother board graphics) draws about 80 watts. While I dont have access to a T605, we do have quite a few similarity equipped HP380s in the server room at the office. For round number back of the envelope calculations I usually figure 400 watts per when adding one to a rack where I have to plan on starting the whole rack up at once. The number and speed of the disk drives seems to matter most, cpu load can cause the fans to speed up and increase consumption. Typical idle consumption is probably around 300watts. I'd love to hear the numbers for a four drive T605 at idle, start up, and under load.

While I'd have to agree that ESX/ESXi is a great way to go for all sorts of reasons, unless you're replacing *many* power hungry machines with one server, you really arent going to save a lot. For comparison, 400 watts will let you run 40 overclocked (266mhz) NSLU2's. There's a whole lot of home automation that can be done with a single NSLU2, imagine what 40 will do for you!

Now OTOH, if you take that dual core desktop and a couple of the slowest 500gb drives you can find, and put VMware server on it (or ESXi if you want to spend the money on a hardware RAID controller) then load that up with a few XP Pro or Linux images and you're likely to still be drawing under 100watts. Of course, each VM wont be all that powerful, but all that really matters is if they are powerful enough for the task at hand.

 

[kwilcox]

I guess I'm gonna have to find out what the T605 is really drawing & compare to what the old machines used. Since the T605 uses quad core Opteron procs and 4 SAS drives, it'll be very energy efficient. In fact, AMD cites this as a reason to use their quad core chips. We use DL380s at work too but ours are mostly based on older dual-core Intel chips and are pretty power hungry. We always dedicate two 30A circuits to our racks for that reason. We do have a DL380 G5 coming in ITNF however.

I guess I had always thought that a power supply drew maximum wattage at all times. Interesting to hear that this isn't the case...

 

[felixrosbergen]

Nope...just as an additional my AMD dual CPU (MP2000 proc's) with 550w PSU only draws 160w or so when idling (measued through KillaWatt). This 160w is however enough to heat
up my small office...

For my HA/Media server i plan to purchase an 430wat high efficiency PSU and Intel E8400 (Core2Duo) based system which will hopefully cut back the power draw even more.

 

[kwilcox]

Well, I stuck a kill-a-watt on the three old machines. Startup spiked it to about 300W but after they settled down the combined machines drew about 200-210W.

Now to find a minute to take down the T605 to insert the kill-a-watt and test. This might not happen itnf so I tested a R805 running ESX in a similar configuration at work and it drew about the same (200W). the R805 is also a dual socket quad core AMD machine but has dual power supplies, 6 Nics, 2 onboard SAS drives in a mirrored configuration and dual FiberChannel HBAs. There's a DRAC card with its own onboard CPU in it as well which drew 30W when the server was powered off. Tested draw was with one power supply unplugged. Having both running increased the draw by 17W. I'm guessing that the T605 will be in the 190W range based on this (no DRAC but 2 additional drives).

So, it looks like its going to be a wash from a power usage perspective. To make out, you'd have to virtualize 4 physical machines, and then it'd be about 60W savings or about $7.00/month.

 

[kwilcox]

Fortress is getting a sensor system upgrade. The hawkeye PIR sensors that form the current network just didn't have the required sensitivity and it affected detection as well as light-out time. What I required was a more professional grade motion sensor to optimize lighting macro accuracy. There are many motion sensors out there but after testing several, the Honeywell (Admeco) 997 seemed to stand out from an overall value vs. capability perspective. I found the best pricing here.

But how to interface a contact closure based sensor that runs on 12vDC to HCA? Well, there's the Global Cache GC100. This nifty device can handle all sorts of input but at $700 for the total solution to control 6 switched devices, it seemed too pricey.

So, why not mod a $10 hawkeye instead? HCA handles W800RF32 traffic with ease and modifying a Hawkeye sensor for contact closure based transmission is simple. The result would be a highly reliable solution for one tenth the cost. Unfortunately, the stock mods out there are based on the Dawn/Dusk circuitry which doesn't have priority from a transmission perspective. It can take a hawkeye up to 4 seconds to respond to this input. So I engineered a different mod to give me the response times I needed. You can find the writeup for this mod here.

I next placed the modified sensors in a central location near the server and powered them with an ATX PC power supply I salvaged from an old PC. These put out 3.3VDC as well as 12VDC. The power supply itself was then plugged into my existing server UPS to provide power outage immunity. All of the modified HawkEye sensors are powered from the 3.3 volt supply and the Honeywell sensors are powered from the 12V supply. I'll never have to worry about changing Hawkeye batteries again

Finally, I ran 4 conductor, 22 Gauge security wire up to the sensors themselves. Here are the results from the first changeover (as a person who enjoys living dangerously, I went straight for the master bedroom sensor as my "Proof of Concept"):

Shot of the Sensor board showing the ATX supply powering Hawkeye and the first Honeywell sensor:

Shot of the Honeywell Sensor installed in the Master Bedroom.

How much better is this? I have bedroom light-out time set to 1 minute now which is half what it used to be. Light-on detection hasn't failed to trigger yet (I had about a 2% failure rate before). No unexpected lights out while the room was occupied either. It all added up to very much higher WAF especially since the sensor installation itself has a much more professional appearance and is less noticable.

 

[drozwood90]

Did you go with these instead of X10 Security sensors due to the packaging?

--Dan

 

[kwilcox]

If you are referring to the Honeywell sensors as my choice then that was partially it. They really are great looking, unobtrusive sensors with high WAF right out of the box. The other portion was that the security sensors also didn't have the sensitivity that I needed. Finally, with this update, every wireless transmitter that forms the sensor network is now located 6' from the W800RF32 antenna. Now there's no lost transmissions any more. Even with BSR's antenna mod, I did have occasional transmission related issues with sensors at the far ends of the house.

If you are referring to using the Hawkeye sensors as modification candidates then it was two things actually. Thing one was price. I've not seen an X10 security sensor for $10 yet. Thing two was the fact that hawkeye schematics are widely available on the web so I was able to design the modification without having to reverse engineer a schematic first.

I'm really only using the Hawkeyes as X10 wireless transmitters now. Do the security sensors retain their programming when you change batteries? If they do, then they would be good candidates for this application as well. I'm sure the modification would be similar. I've got UPS backup that would run everything for about 4 hours currently. My strategy for an extended power outage would be to either put batteries in the sensors while they were still powered up from the UPS or just resign myself to re-programming everything once the power was restored. Everything's labeled so it wouldn't be a big deal.

 

[drozwood90]

If you are referring to the Honeywell sensors as my choice then that was partially it. They really are great looking, unobtrusive sensors with high WAF right out of the box. The other portion was that the security sensors also didn't have the sensitivity that I needed. Finally, with this update, every wireless transmitter that forms the sensor network is now located 6' from the W800RF32 antenna. Now there's no lost transmissions any more. Even with BSR's antenna mod, I did have occasional transmission related issues with sensors at the far ends of the house.

If you are referring to using the Hawkeye sensors as modification candidates then it was two things actually. Thing one was price. I've not seen an X10 security sensor for $10 yet. Thing two was the fact that hawkeye schematics are widely available on the web so I was able to design the modification without having to reverse engineer a schematic first.

I should have been more clear with the question...I meant, instead of using the Honeywell sensors...use the security sensors...but thenk you for answering BOTH questions.

I'm really only using the Hawkeyes as X10 wireless transmitters now. Do the security sensors retain their programming when you change batteries?

Yes, they will retain their programming for...well...as long as it takes me to "normally" change the batteries. I have only had to change them once...and so far no lost codes. I think I've heard upwards of 10 minutes...although I've not personally found this to be a good estimate of time. I've not really experimented to determine how long. However, in the few minutes it takes ME to change the batteries, they do NOT loose their codes.

I would really suggest you look into using DS10As. They are security (so you can in theory have 65K of them) and you won't get 100's of transmissions. Modification is even easier then the Hawkeye, as they accept a relay closure as a trigger. If you need, I have a copy of a few links that brings the price of these units below the $10 range (I think I had them for $3.33 each). I have changed the batteries on these as well...they tend to NOT loose their codes for a few minutes as well. I really like the way you have done this!

If they do, then they would be good candidates for this application as well. I'm sure the modification would be similar. I've got UPS backup that would run everything for about 4 hours currently. My strategy for an extended power outage would be to either put batteries in the sensors while they were still powered up from the UPS or just resign myself to re-programming everything once the power was restored. Everything's labeled so it wouldn't be a big deal.

I am SO interested in your setup, as what you did / am doing is exactly what I have been striving for! My main concern has been the bedrooms. How do you UNOBTRUSIVELY put a sensor in there without loosing sensitivity?

For my house, the bottom floor is kind of open...so one can't tell if someone is in the hall, or in the front reading room...that is with a "360" sensor as you have linked. It's also more difficult for me to get wires strung to the ceiling of those rooms...as there is another floor above! What's even more fun

Personally, my top floor would be perfect for getting at the bedrooms and the hallway done with these Honeywell sensors. I can drop it in through the ceiling, and run the wires down into the basement through my future tube.

Then all the bedrooms should only need that ONE sensor, as well as the window sensor.

For each room, do you have a way of turning the light "permanently" off at night? How about if someone gets up in the middle of the night? Does their room light come on automatically at all?

For my bottom floor, I think I'm going to stick to the X10 security sensors as even though they are "big" they can be "directed" and do not need wires. The other issue I have, is to run a wire from the basement to each of the rooms in the bottom floor would be incredibly difficult, as the best place to access the ceiling in that floor is "left to right" (facing the front of the house), but the floor joists run "front to back"...not something I wish to attempt.

I have a few more "offline" questions I'd like to ask. I'll PM you!

--Dan

 

[kwilcox]

Personally, my top floor would be perfect for getting at the bedrooms and the hallway done with these Honeywell sensors. I can drop it in through the ceiling, and run the wires down into the basement through my future tube.

Then all the bedrooms should only need that ONE sensor, as well as the window sensor.

Same here. For the main floor, its trivial to run PELV wiring in the attic.

For each room, do you have a way of turning the light "permanently" off at night? How about if someone gets up in the middle of the night? Does their room light come on automatically at all?

Yes. Turning off the UPB wallswitch for a room also disables the macro that controls automatic lighting. Here's how I do it:

Each room has a simple macro that triggers on a status update from the associated wall switch then sets the flag called "<room> Manual Mode" accordingly: if the light is turned off then "<room> manual mode"=True else it is False. The "go to bed" macro sets all of these flags to True. Conversely, all flags are reset at 12PM each day.

Lighting macros are always triggered by an x10 On from the associated room sensor. Essentially, in their most simplistic form, they do this:

Start:
<room>manual mode? yes=exit
dark outside? no=goto wait
light on? yes=goto wait
house sleeping? yes=activate low level lighting else activate high level lighting
wait:
wait 1 minute
light on? yes=turn light off
exit

The macros are set to restart if the trigger occurs while they are running which means that a full minute must elapse without motion input before they will turn off the light. To improve response time, the light on state as maintained by HCA is used. No actual status commands are sent.

For my bottom floor, I think I'm going to stick to the X10 security sensors as even though they are "big" they can be "directed" and do not need wires. The other issue I have, is to run a wire from the basement to each of the rooms in the bottom floor would be incredibly difficult, as the best place to access the ceiling in that floor is "left to right" (facing the front of the house), but the floor joists run "front to back"...not something I wish to attempt.

Same issue here almost. I will have to ponder whether the difficulty involved with running PELV is worth it in the lower level.

 

[kwilcox]

Forgot one answer:

My main concern has been the bedrooms. How do you UNOBTRUSIVELY put a sensor in there without loosing sensitivity?

I should include a pic of my "sensor jig". Essentially its an 8 foot tall stand that places a Honeywell sensor near the ceiling, has its own modified hawkeye sensor and 12vDC supply. To use it, I program the jig's hawkeye to the room code then disable the room's original hawkeye sensor. I can then experiment with placement locations to maximize coverage. When everything's right, I just put a pencil mark on the ceiling, check for studs then drill baby drill!

Warning: an 8 foot tall 2x4 sensor jig with wires and zener diodes hanging off it while plugged into a wall socket and placed by the walk-in closet has incredibly low WAF!

 

[beelzerob]

Heheh...I like the creative use of the ATX power supply. I have about 5 of those just laying around..I'll have to remember that before I go out and buy a 12v power supply for something.

 

[kwilcox]

Heh... don't we all...

Here's the pinouts:

Just short green (PS_ON) to any ground to power it up. I snipped off the green and the adjacent black then stripped/twisted together and used a wire cap. Fan won't run unless its getting hot under load.

 

[kwilcox]

Well, the WAF on the Honeywell sensors is not perfect. The mechanical reed relays make a soft chirping sound as they turn off and on. This is especially noticeable in the master bathroom which has harder acoustics and is smaller than any of the bedrooms. I believe I can correct this with minor circuit modifications though. Essentially I'll simply replace the relay with an SSR for silent operation. The Clare CPC1219 looks ideal for this purpose.

There's always something...