|
>> this doc states that a linux machine on 100Mbit network will top out
after
>> 8000 pps. mine traffic already produces more then that, separated across
>> several machines. if i convers these several machines to be realservers,
>> then the director will get more the 8000 pps . is the doc above still
>> relevant ?
>
>That probably refers to tests runs on older, slower hardware.
>As far as I know there are no limitations explictly set in the code.
I'm just about done with a setup of linux-routers using Zebra to replace our
Cisco gear. Non NAPI drivers in 2.4.x seem to handle 33k pps with packet
size of < 200 bytes. With NAPI (sort of polling mode in linux) I have
heard you can get 140k+ PPS on 100mbit. If you are worried about your box
freezing up from too much pps then use those drivers, they are stable.
I definitely agree that switching to LVS-DR rather than upgrading your
entire network setup is a good option. Also if I were you I'd get 2 new
boxes and get this 2.4.latest + NAPI + LVS-DR/HA setup working on 2 new
boxes rather mess with your existing setup.
Hope that helps, and mail me if you need NAPI urls.
Peter
|
