Hall Effect for outlet strips

Certainly with insulation type and thickness compliant with UL rules. Maybe even listed, for a couple more bucks in quantity. I've bought

John

So on the NANOG list, there's a thread on where to find a rack-mount outlet strip with per-outlet measurement, for bill-back purposes in a server hotel environment.

The data so far is grim:

{baytech}
> I had moderate success with this suggestion. Their technical support said
> the only product they had that does this is the 4 outlet RPC5 or RPC6
> (ethernet version vs serial version). Unfortunately, it costs $644 each
> (lowest price I've found so far) and accomplishes it's individual
> monitoring by replicating power in and power out plus an ethernet port 4
> times. Still, if it's the only one out there I guess they win (although
> at $150 per outlet, ouch, that goes over my $4000 budget for this).
>
>

formatting link

And it's not UL-approved.

So I was thinking; is there a Hall-effect package that could cheaply tell you raw amps at least on each of n outlets?

This may have come up before, but:

Buy a regular outlet strip and n Kill-A-Watt single-plug watthour meters, about US$35 each. These meters don't have data output, but this can be somewhat fixed by 1 to n cheap webcams and creating a web page for someone in the billing department to look at this month's reading, last month's reading, and doing the math. For more automation, there is a single-plug watthour meter with serial port, the Watts Up Pro. About US$125 each.

I think some UPSes (APC Smart-UPS?) report the current load, even when they're not on battery. Maybe you could buy a cheap, small UPS for each server, and use its built-in monitoring. You're not after the UPS functionality - that is still provided by the big UPS and generator for the whole data center.

If the hotel owner can run their own code on the servers, you might be able to do it by proxy. Temporarily install a watthour meter and note its reading for the server in each of three states: idle, running all the disks ('find /'?) only, and running 100% CPU (Seti@Home?) on all CPUs only. Then use something like MRTG to collect stats from each machine and compute electrical usage based on the disk and CPU load and the measured energy consumption.

I've only bought co-lo space, and not sold it, but I wonder: is price so important and margin so low that the ability to charge a customer less if they use less electricity is that important?

Digi-Key sells a Melexis MLX90215 Hall-effect IC that looks like it might work for $11/1 and $7/50. There is a page on the data sheet about using it for current sensing. I don't know anything about it other than it exists.

Matt Roberds

That is not how it works, you have to sample the momentary values and multiply them, process, then embed an ethernet controller etc. Not a trivial task. I also would use just current transformers, hall sensors (Lem makes them) are even more expensive and require extensive analog conditioning. So with a PIC and a CT per outlet you could probably come along. But you will not have the skills, if you don't know about phase angle. Go buy that ready made outlet!

....

So skipping the ugly kludge; you're talking $125/port.

Look; space is money. The wasted space of the UPS's is just that...

Does not fit the requirements.. They want data, not SWAGs...

When you consider each watt bites you twice -- once for the box, again for the HVAC to remove the heat...

UL-satifying isolation? I suspect not...

Hey you talk like a computer tech. Have you ever ask yourself, what should this expensive investment be good for? The guest will only hook up his laptop or his cellphone charger. One hot shower or 10 min. of hairdryer will cost more than 10h of all that equipment. And what can the hotel owner charge? 1$ per kWh? Then this investment will never pay off.

This won't appease the techies on NANOG, but this problem doesn't scream for a technical solution.

Servers shouldn't vary their consumption much, so why meter? Set the billing rate based on either the rating plate of the power supply, or a one-time reading of the server before it's installed. Both are easy one-time checks, and they're defined values that both parties can plan around. (And yes, they both have weaknesses that could be exploited, but significant enough to need a much more complex solution?)

The facility needs this data to size the circuit anyway, and all the facilities I know charge by the circuit anyway to recover power and cooling costs (widely varying, but ballpark $200/20A per month).

Richard

I should have thought that a simple hook-up fee would more than cover any energy usage that was physically possible/probable, if all you want is another frigging item line on the hotel user's bill.

RL

Sir:

I fear you misunderstand the basic issue.

A "server hotel" is a term for a dedicated facility that houses servers, hundreds of servers, and nothing else but same. There are no hot showers, hairdryers, or even windows. [1] And the only food MIGHT be a junk food vending machine in the office area, if there even is one. {Most are unmanned...} If there's someone there at night, they are most likely frantically trying to get their customer's server back on-line.

And depending on the lease terms, the landlord does ndeed charge back for every KWH used. The question was and is how to easily log same.

snip.....

I guess he's not the only one mislead by your use of the term "server hotel".

Perhaps you might consider a portable 'mousetrap'-type monitoring of slow-peak energy usage - as this is what affects your hardware sizing and infrastructure costs, rather than actual consumption. The two readings are likely corelated.

A self-powered CT with a long time constant average current measurement, a (latching?) presetable threshold and indicators of in/out of range. The user pays for the in-range threshold setting applied, with out-of range penalties. Very manual monitoring and all hardware - I know - but low-tech and also potentially log-able at 'one' link per monitor. How the info is handled.......

The usual problem with VAR relationships occur, when current is the unit of measure, but there would always be bitching, anyways.

Alternately, simply charge by the input power label rating of plugged-in equipment. A calculation of total plant usage vs total label ratings would give a ballpark cost per label-watt-hour in any billing period. Utility rates are a variable. Accuracy is probably unimportant - you just need an administrative procedure that produces cashflow, supposedly for energy delivery cost compensation.

RL

Yep. Specs stated or implied in your first post were: rack-mount outlet strip, per-outlet power measurement, UL approved, price to beat is $161/port. In the absence of other data, I figured kludginess vs. ~$80/port savings was a good trade-off.

The APC 350 VA "fat power strip" UPSes can be had for $30 each. $95/port vs. space.

But they don't want to pay very much for the data.

And, AFAIK, this is normally reflected in the pricing of the rack space

- when you pay for a 15A or 20A circuit, you're paying for the electricity to bring the heat in and the cooling to take it back out.

Again, it would be really interesting to see the business case for this.

Matt Roberds

So you *are* a computer tech. I'm not that familiar with IT slang, only know server farms. THX for the explanation. BTW it's kWh not KWH. A pity you are not familiar with circuit design, because in your case where hundreds of monitors are needed, it might be useful to develop a small gadget with a PIC that does accumulate the consumed kWh and reads them out via ethernet. And such a gadget will even sell to other similar businesses, so there might be a nice-market for the power/work monitor. I would guess the hardware costs below 25$ for moderate runs.

I bought a W/kWh/VA/current/voltage/PF meter for $32:

(sorry for the wrap)

Not at all; I'm a EE. I just do not now do the kind of design/build work that many others here do. (I've done some such work in decades past....)

Well, one basic rule is "don't reinvent the wheel..." so I asked here to see what, if anything, was available already. And if you think you can deliver such a gadget, UL/etc certified, for $25/port; I can point you to lots of potential customers.

That looks like a reboxed "Kill-o-Watt" meter; they are nifty -- I have one on my fridge right now. But it lacks a Ethernet port....

-- A host is a host from coast to snipped-for-privacy@nrk.com & no one will talk to a host that's close........[v].(301) 56-LINUX Unless the host (that isn't close).........................pob 1433 is busy, hung or dead....................................20915-1433

I bought it because it was nifty and to measre the power consumptioni of my computer gear. So far all I've found is that my bedroom AC draws

Oops. Forgot that requirement. How about X10 security cameras, with... ;-)

You have too much time on your hands. ;-)

Some things have no answers. I have to move the 'fridge, range, and island out to the front porch tomorrow (good thing we have no crappy HOA), so maybe I'll have time to play with it. ;-)

So I have one on my fridge. The compressor is off, so the only draw is maybe the defrost timer?

I open the door, the light goes on, and the PF gets *worse*... WHAT? Then, later, I try again, and the PF improves to near unity.

Still pondering...

-- A host is a host from coast to snipped-for-privacy@nrk.com & no one will talk to a host that's close........[v].(301) 56-LINUX Unless the host (that isn't close).........................pob 1433 is busy, hung or dead....................................20915-1433