search

cpignata / eimpact-icmp

4 stars 0 forks source link

Add component throughput #14

Closed mwelzl closed 6 months ago

mwelzl commented 8 months ago

We have node power, node throughput.

We have component power, but not component throughput.

Not only is this inconsistent but the same logic applies: whatever can be calculated for the node using node power and node throughput can be calculated for a component if we have both.

cpignata commented 8 months ago

What if components do not have throughout, like a CPU or a controller card?

On Wed, Mar 6, 2024 at 6:22 AM Michael Welzl @.***> wrote:

We have node power, node throughput.

We have component power, but not component throughput.

Not only is this inconsistent but the same logic applies: whatever can be calculated for the node using node power and node throughput can be calculated for a component if we have both.

— Reply to this email directly, view it on GitHub https://github.com/cpignata/eimpact-icmp/issues/14, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOP5KZ5W475R4QJBUSQ7ITYW336VAVCNFSM6AAAAABEI3RWZ6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGE3TCMRWHE3TINQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

mwelzl commented 8 months ago

Then we need to return an error message. And, that's the same about your per-component power measurements, isn't it? Sure everything must be powered but power readings may not be available. Should we open a new issue on errors?

cpignata commented 8 months ago

Might not be an error per se. Let’s think through the model. Maybe a bandwidth of some number or flag means not applicable. Thumb typed by Carlos Pignataro.Excuze typofraphicak errowsOn Mar 7, 2024, at 01:52, Michael Welzl @.***> wrote: Then we need to return an error message. And, that's the same about your per-component power measurements, isn't it? Sure everything must be powered but power readings may not be available. Should we open a new issue on errors?

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: @.***>

mwelzl commented 8 months ago

Yeah - some such message. I agree, "not applicable" sounds better than error.

mwelzl commented 8 months ago

or "not available". That's maybe better - it's broader. For some components, power or throughput measurements may be "applicable" but still not "available". This gives us one notification style that would fit for everything.

jmparikh commented 8 months ago

I like the idea of setting a specific flag (like a sequence of bits) or just setting the bandwidth to a reversed number to convey "not applicable/available".

cpignata commented 6 months ago

This is awesome!!! @mwelzl , I'd just add a reference to rfc4122.