Network power usage monitoring - Streaming Telemetry?

[DJW487]

I am interested in monitoring the total power consumption of the 700+ switches in our network.

We have over 1500 APs plus the usual assortment of IP phones and other things running over PoE.

The idea is that I can get a baseline for the overall consumption, even better if I can break it down into sites (50+ sites) and graph it (I already have Grafana for other things so bonus if it can natively integrate).

I will then look to start implementing things like power efficient ethernet and some of the newer power profiles for AP's which allow them to ramp down and up during quiet/busy times and track the difference in savings to present to the company.

I'm sure this could be achieved by conventional SNMP monitoring, but I'd be keen to know if anyone has set it up on 9200/9300 switches using telemetry streaming?

Or does DNA/Catalyst centre already receive this info and I can just tap into it somehow through API?

[Leo Laohoo]

How granular do you want?  Per port?  Per switch?  Per power supply? 

How long does the the record hold, 30-, 60-, 90-days?  1-, 2-years or longer?

Per minute, per hour, per day?

[DJW487]

Hi Leo!
I think per switch/stack is probably as low as I would need the granularity.

I generally aim for 90 days of data for my graph retention and then I could do a report per quarter.

I would like to see per hour to begin with to get an understanding of how things change throughout the day.

[Ben Walters]

It seems like the PoE monitoring is a bit of a mess in DNAC right now, we have telemetry enabled on a couple cat9300s and the PoE values are not being reported correctly. 

Most of the time what DNAC shows for the consumed watts is lower than the actual consumption when you run show power inline on the switch.

The API is also problematic. You can see the overall PoE consumption for a switch and it is accurate but it only shows switch #1 in the stack when you run a GET to /dna/intent/api/v1/network-device/${deviceUuid}/poe. I can't seem to find a way to pull any other switches in a given stack since they all seem to share the same device ID in DNAC. 

Our DNAC version is 2.3.7.5-70434. Based on my tests it looks like the PoE monitoring is only partially functional within DNAC.

[Leo Laohoo]

I use a different product which uses SNMP to continuously poll our network appliances.  This product does not discriminate -- If it has an IP address, regardless it's age, it will ping-n-poll every 60 seconds.

On a switch-level (vs per port), this is what it looks like every 60 seconds, 30 minutes, 24 hours.  And for "trending", I can see what it is like 1 year, 2 years or more.  

Note:  The graphs are not limited to what I've just enumerated.  I can see 2 minutes, 5 minutes, 7 minutes, etc.