I did not say anything about what is “fair”. I am, however, concerned with useful and transparent measurements so our customers can make informed decisions.

Max draw is primarily useful for the electrician who is provisioning power to the switch–regardless of working load, you need to provision for the plate rating. However, when trying to figure out cooling capacity, working off the plate rating generally leads to inefficient use of budget and over-built capacity or cooling capacity in the wrong places, especially when we get into modular switches. For example, our Catalyst comes with 6000W power supplies but if you look on our DCAP best practices web site (http://www.cisco.com/en/US/netsol/ns758/networking_solutions_sub_program_home.html cisco.com acct req) you will see, the 6500 serving as an agg switch draws 2273W. Now, you still have to provision 6000W, but it is helpful to know just how close you are to your to your power capacity and cooling capacity. It is also helpful not spending money on cooling that may not be needed. BTW, as max load, per their respective data sheets, there is only 10W difference between the 4548GT and the Catalyst 3750G-48TS.

At the other end of the spectrum, I still do not believe that testing a switch at “idle” (no connections, no load) is a useful measure, because that is not a normal operating state for a switch. Testing at low-load, is a more useful metric to give customers a better understanding of the operational range of a switch (i.e. low-load to max draw). In their tests, Miercom measured each switch in the test group at “idle”, at 5% uplink load and 100% uplink load with both copper and optical interfaces, which is probably a more realistic usage scenario, especially since we are talking about switches that will likely be deployed it the access layer.

As an example, when Miercom tested the Nortel 5510-48T, the idle/no-connection draw was 80W, but at 5% uplink load with L2 traffic, the draw went to ~125W, or about 50% increase. BTW, this behavior was not unique to the Nortel switch.

As I said in my initial response, we ultimately need a set of standardized test and testing protocols that we, as manufacturers, can run, so that customers can make informed decisions.

Omar

But no need for common sense, I’m sure in the near future those tests with simulated traffic will come out as well.

I am not sure what your comment means, since my comments are based on Nortel’s published documents. My original point was that measuring switches while they are not doing anything does seem all that helpful to customers. It would by like trying to extrapolate gas mileage based on how fast a car idles. If you have third-party testing of power consumption for Nortel switches under load, that would certainly be helpful in continuing the conversation in a meaningful direction.

Omar

The actual NEEC has 30+ inputs to it, making it a very real-world analysis tool. The page you link to is an overly simplified mock-up to support an advertising campaign (since we all know the average reader has a 2 minute attention span).

You can see actual videos of the NEEC on YouTube at the below links, and I’m also happy to show you the real calculator in action.

The NEEC will be publicly available online in a month or two (quite the conversion process to take an analysis tool this complicated from an application to a webpage).

Here’s a specifc example for a customer in Japan (bad audio): http://youtube.com/watch?v=40SvriCCQPo

About 1 minute into this video are several shots of the NEEC: http://youtube.com/watch?v=0agxNrONTr0&feature=user

Totally agree–I think the best part of all this is that the conversations are being had and consciousness is being raised. In the long run, I think we, as an industry, will figure out what kinds of metrics make sense, if for no other reason than customers will ultimately demand meaningful data that allows them to make concrete discussions.

Omar Sultan
Cisco Systems