On Mon, 28 Feb 2000, Dennis wrote:
> vmstat -i ; sleep 5; vmstats -i
>
> Just subtract and divide by 5...most 4th graders can divide by 5 2nd
> graders with a calculator.
Denis, why dont you take a deep breath and relax. Go and get a drink or
something.
Why are you so angry?
> Its not exact, but pretty close. The same can
> easily be done with proc/interrupts. Its not a window, its a count.
>
> Why are you arguing these silly, ridiculous points so easily shown? Are you
> trying to prove that the 110k number is not possible or not 100% accurate?
> What exactly is your point?
I am saying the 110K interupts/sec sounds very high and you might be
inaccurate. But i'd really like to know, honestly.
If you measurement window is 5 seconds, then what i am hearing is you have
a sustained rate of 110K-interupts per second for that 5 second
period. What i am arguing is that you might actually have had peaks of
200K-interupts per second in there.
Thats where your inacurracy is.
How is you device generating interupts? Per collected complete packet or
per 40 byte data that you used in your "math"?
> >
> >> Lets see, 90Mb/s / (40 * 8) = over 281,000pps max per HSSI, and I said dual
> >> HSSIs so thats 560,000pps possible.
> >>
> >
> >Ok so 40 bytes are just IP headers; no framers needed because T1
> >technology is synchronous.
> >But i was under the impression that a T3 would probably give you a max of
> >45Mbps
>
> You really dont know how big an IP header is? Its 20 bytes last I heard.
>
True; for some reason i added the TCP headers as well ;->
Now it could be much higher than 20 bytes when you start adding options;->
But lets not go there.
> "synchronous" has nothing to do with it..actually tcp packets will be 44
> bytes minimum but a PING for example with 1 byte of data could be only 33
> bytes (a 4 byte encapsulation)...40 was just an example.
>
I was doing my earlier calculation based on 64 byte packets that are
the minimal allowed on ethernet.
Now you are telling me 40 bytes was just an example? You based your claim
on 40 byte packets.
What is the minimum size packet allowed for your hardware?
> Serial lines are full-duplex, so 90Mb/s can be transfered at once.
You are still limited by the serialization speed.
What do you use for framers? PPP? If you are depending on the DS0 time
slots, then thats why i am making the arguement that you are dependent
on "synchronous" activity. I could be wrong of course.
>
> >Denis, you techno-ruffian God! Please have it in your heart to forgive us
> >mere mortals when we sin.
>
> But why do you continue to argue subjects on which you are so poorly
> versed? Why must you contradict everything I say?
I am not. You are being a little too sensitive. Why not attack the
message i post instead of me? Your postings are normally littered
with Ad Homimems. Just look at the the above quote from you.
Get out of your marketing suit because that would help.
cheers,
jamal
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
