Re: PPP speeds; "nice"-ness
In article <cistron.19990826233228.G24322@willamette.edu>,
Seth R Arnold <sarnold@willamette.edu> wrote:
>On Fri, Aug 27, 1999 at 07:01:14AM +1000, Alan Eugene Davis wrote:
>>
>> This prompts me to ask, is there a concept of "nice"ness for TCP/IP
>> connections?
No it doesn't exist.
>> How does apt arrange to have a priority, as apparently
>> it does, and arrange to have good speeds when other processes don't?
TCP slow start. TCP starts out slowly, then goes faster and faster.
So long transfers eventually run at top speed while short lived connections
don't.
>As for how apt grabs most of the bandwidth, I don't know -- how to fix it, I
>have one suggestion -- the tcp/ip traffic shapers of the 2.2.x kernels. I
>don't know whether they work or not, or if they will go with ppp
>connections, or if it will help a system that gets 3kps tops -- but it is an
>idea. :)
That will not work.
>The traffic shapers (from all that I know in the make menuconfig help..)
>should allow you to limit bandwidth usage perhost -- probably not per
>command.
No, that work only for outgoing data. You cannot control how fast the
other host sends you data, unless you configure a traffic shaper on
the sending host.
Mike.
--
... somehow I have a feeling the hurting hasn't even begun yet
-- Bill, "The Terrible Thunderlizards"
Reply to: