Faster Network & Save Money?

[Robbie]

A few external prompts to reorganise my home network - my 10 GbE switch failed, I installed a larger rack, NASA left home (aka my daughter - a subnet in her own right) and I had moved a couple of wifi APs, enabling me to delete / repurpose an AP.

First task - find every item on the network and ask 'do we really need it', no really. This resulted in 5 wifi clients being pulled outright and running cable to an additional client that does have an RJ45 port - I already have a policy of running everything that does not move on ethernet to keep wifi just for mobile clients or guests. I then pulled over a dozen ethernet clients that served little purpose and, as one was a lagged NAS, it freed an even greater number of network ports. As all the core elements of my network get a DHCP reservation / static address the cull brought me down from around 80 static addresses to a little over 50:

A rework of my 24-port switch and a 5-port switch achieved a greater port density than previously; it also host my 10 GbE clients at just 2.5 GbE for now:

The port rework along with the cull of wired ethernet devices left my 48-port switch in this unusual state:

Hurrah!

Along with the removal of a temporary 10 GbE switch and with a second 5-port switch released from duty I can now run all the 'steady state' network clients on just 2 switches. The UniFi US-48 remains installed for any development/testing or stepping in if/when NASA returns home.

As for achievements - well no part of my network is deeper than 2 layers, so reduced latency (apparently) and more simple switching & management. The tweaked wifi AP positions gives me an equal minimum of 4 wifi clients each for the 4 APs, rather than a cluster around 2 of the 5 APs. The new topology is way flatter than before (or way taller in the unfriendly UniFi topology map):

The sprawling network has been contained!

On the saving money side the UPS power draw has gone from pulling 180 W steady-state, to a miserly draw of around 70 W - that should save the odd tree or something for the items running 24/7. The always-on 70 W or so powers an EdgeRouter 4, 2x UniFi PoE switches, Mac mini M1, RS1221+, NTP server, 12v rack accessories, 2 sensors and the standby power of an RS217 and RS819. I think that is an impressively low number!

Additionally the RS819 comes on for a few hours a week for backups and the RS217 is remotely triggered only when required for testing. The UniFi 48-port switch does not have a proper standby mode though (you can command it via CLI to sleep, where it still pulls around 15 W, but you cannot wake it back up again without a power-cycle). I've put it on a switched socket on the UPS (Group 2 above) to remotely power it on/off via a browser, with zero quiescent power draw.

Aside from 2 bits of future work (1 x 5-port switch to come off PoE power and 1 x AP to go on to PoE for emergency use) I should be good for a while. It is the first time I have felt better with less kit on the network but it all runs smoother, produces less noise and heat and should save a few pennies along the way.

Anyone else up for a network clear-out?

"My name is Rob [Hi Rob] and I am a networkaholic. I have been free of superfluous network items for 2 days..."

️

 

[jeyare]

Nice job, Robbie.

Two years ago, when I completely switched to the Unifi platform (gtw, swch, APs), I similarly redesigned the topology. However, I invested in the PRO48PoE switch because it gave me more freedom in the future (confirmed now).
I know very well the scheduled power by UPS for the switch (night rest, when all NASes finished their backup processes). The Unifi controller is running in a container, then no need to worry about the graceful shutdown.

From the EATON screenshot, it is not shown Apparent power in VA (there is just Real power in Watts).
Follow the Eaton white paper for the 5P 650i Rack1U model, there is 0.65 PF only (my Legrand is about 0.99 = more compensation there). So your power consumption is about 103.1VA = 67W x 0.65. What is still OK for such equipped rack. Better to see the Load timetable (based on common load from the mentioned HWs).
Well done.

 

[Robbie]

@jeyare - In an ideal world perhaps I could achieve such high PF ratios but at my load levels I get quite a bit lower than the book figures suggest.

During my pruning of the network a couple of days ago my PF was only slightly above 0.6 at 37% load.

Now I am down at 26% load my PF this morning was just 0.39:

Time (UTC)	Output Voltage (V)	Output Frequency (Hz)	Output Current (A)	Output Apparent Power (VA)	Output Active Power (W)	Output Power Factor	Output Percent Load (%)	Battery Capacity (%)	Battery Remaining Time (s)
2021-09-14T12:09:06.00	243.8	50.0	1.0	244.0	147.0	0.6	37.0	100.0	770
2021-09-16T10:38:17.00	245.7	49.9	0.7	172.0	67.0	0.4	26.0	100.0	1748

Ah well, at least the battery runtime has increased significantly.

️

 

[jeyare]

Re Battery Remaining Time:
I can't get it (challenge); follow the data in the table, because when:

Row A. Output AP is 244 VA = 37% of Load ... 100% = 659VA
Row B. Output AP is 172 VA = 26% of Load ... 100% = 662VA
Note: the small rounding difference is negligible, then both values are OK and near to the official 650VA/420W capacity for this model (0.65 PF);

but

for the A. it is about 770s of Remaining time = 659VA / 770 = 0.86VA consumed per 1 sec.
for the B. it is about 1748s of Remaining time = 662VA / 1748 = 0.38VA consumed per 1 sec.
what is strange gap
because
the Remaining time for B. can't be 127% higher than A. (1748 - 770 = 978 / 770 sec = 127%), when the output AP of B. is just -30% less than A. (172-244 = -72 / 244 = -30%)

I can't see there the PF penalty for the B.

More accurate is the value 1000 seconds for the B. (follow the math in A.)

From the physics point of view, the battery capacity utilization time can be increased only with the use of a higher PF (which in practice is corrected mainly in Data Centers).

Checkpoint by Official Eaton Runtime chart (attached) for this model:
147W load = approx. 14 minutes runtime = 840 secons ... 147W / 0.65 = 226 VA; what is -7.3% less than your AP in A.
67W load = approx. 31 minutes runtime = 1860 secons ... 67W / 0.65 = 103 VA; what is -40.1% less than your AP in B. ... then the resulting time must be shorter = 1860 - 40.1% = 1114 seconds ... near of my math about 1000 seconds, too far from 1748 seconds defined in B.

Eaton 5P UPS

Eaton 5P tower og 1U rack UPS (650-1550 VA) til rackskabe, serverrum og små datacentre. LCD interface og forberedt til netværk. | Overblik

powerquality.eaton.com

Conclusion:
The reason for the differences between EATON white paper (or their calc. prediction) and reality is based on low-end PF compensation used in this UPS, which can cut almost 40% of the battery capacity, as was defined here. Therefore, it pays to consider diff. prices of a model with a similar capacity but higher PF. The manufacturers of the devices are very happy to play with consumption numbers in Watts (W). More accurate is the consumption calculated in VA. The same value in W can have different (higher) values in VA. It depends on PF inside. One of the reasons why I hate the EU Energy Labels - it is total misleading for consumers.
Be careful.
No doubt, it is everyone's own choice. Take it just like a consideration.


Note:
I used a simplified calculation with a linear decrease for both cases, including the use of 100% battery capacity. What will never happen in practice.

 

[Robbie]

No doubt Peukert's Law is hiding in there somewhere.

️

 

[Robbie]

The manufacturers of the devices are very happy to play with consumption numbers in Watts (W). More accurate is the consumption calculated in VA. The same value in W can have different (higher) values in VA. It depends on PF inside. One of the reasons why I hate the EU Energy Labels - it is total misleading for consumers.

With residential consumers paying for their electricity in watts I don't see anything wrong in the EU advocating for the use of watts on all consumer labels. Yes, things are different for industrial users but they also have the means to increase efficiency or have regulations that mandate measures to control PF.

 

[jeyare]

Robbie, it was my idea about the regulation also, till I have seen the Regulation sources.
Last valid is Regulation 2017/1369/EU from 1 August 2017.

There isn’t mentioned Active/Reactive/Apparent power or Power factor. Nothing like this.

Main indicator of the energy efficiency is a “virtual” energy efficiency index EEI, defined by annual power consumption relative to a reference consumption that is based on the storage volume and the type of appliance (e.g. for fridge).
The label contains annual energy consumption in kWh (nothing to do with the EEI).
For washing machine they use for a reference mix of the washing cycles, a weighted mix consisting of 42% full-load cycles at 60 °C, 29% partial-load cycles at 60 °C, and 29% partial-load cycles at 40 °C. The washing performance is not mentioned any more, since all washing machines must reach class A anyway. For a 6-kg machine, an EEI of 100 is equivalent to 334 kWh per year, or 1.52 kWh per cycle.
And so on.

And little science here:

ShieldSquare Captcha

-- post merged: 17. Sep 2021 --

P.S.: I’m an author of many analyses of Reactive power impact in E2E Power distribution environment, based on on data research. Include impact to GHG increase.
The Reactive power is necessary for energising the magnetic circuitry. No doubt. However, mostly it is the “side effect” of low-end equipped electric gears. Because efficiency is costly. And people like …

 

[Robbie]

As said, unless regular consumers are billed in VA there will be zero interest in anything other than kWh. In my country we pay a daily charge plus usage in kWh. The standing charge includes infrastructure and additional overheads, which includes such thinks as PF.

In my country (UK) there we are only just seeing a growing confusion over why single phase remains the default for all residential power networks when, in reality, we really needed to switch to 3-phase years ago.

The push to EV, Solar and Solar Storage will have to be addressed by 3-phase at some point. Some of our European neighbours already express surprise at our continued use of single-phase AC.

 

[jeyare]

The standing charge includes infrastructure and additional overheads, which includes such thinks as PF.

Nothing is for free. Reactive power is calculated for all households in the "standing charge" (hidden item). But then the problem is that if you have a well-tuned household operation = high enough processing efficiency, you are penalized by others who do not care.
Sometimes even compensation is not necessary, because the inductive ratio is the same as the capacitive one and it compensates for itself. However, the inductive ratio prevails because the consumption of devices containing a el. motor predominates in households.

In my country (UK) there we are only just seeing a growing confusion over why single phase remains the default for all residential power networks when, in reality, we really needed to switch to 3-phase years ago.

Of course, because 3-phase consumption is much more balanced than single-phase. Here in SVK, it is a common attitude.

The push to EV, Solar and Solar Storage will have to be addressed by 3-phase at some point. Some of our European neighbours already express surprise at our continued use of single-phase AC.

Especially if the UK wants to reach its goal of moving 33M registered passenger cars to EVs by 2035. You have to expect that the household consumption ratio will increase substantially from today's 33%.