Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Mailing Lists -> Oracle-L -> Re: LAN N/w thruput, Basic Qs - Off-Topic
Just to develop that at bit further, allowing your switches / NICs to
autodetect /
auto-negotiate can do a lot worse than that... I've personally seen
situations where
a 100Mbit NIC was "maxing out" at somewhere between 5 and 10 Mbit/s until
autodect was disabled. Granted, that was a *very* old PC running Windows,
but I am led to believe that this situation is surprisingly common.
I am told by people who know much more than me (not that this says much)
that it is a good practice to hard-code link speeds and duplex settings
rather
than letting your NIC try to guess. The way I see it, even if this
advice is
nonesense it can't really hurt -- as long as you get the settings
right. ;-)
As for actual network bandwidth, don't be fooled by "nominal" bit rates. 100Mbit/s (or whatever) is a "theoretical" maximum. Token based networks can often come close to this nominal maximum, but most other networks where collisions are possible (e.g., ethernet) are likely to peak somewhere between 60% and 80% of the nominal maximum.
'Course, I'm *far* from being a networking expert. It's been years
since I've
been allowed into a wiring closet (and I don't think *anybody* has ever
allowed me into one voluntarily).
Cheers,
-- Mark
Thomas Fox wrote:
>Sriram,
>
>You are correct in your answer.
>
>Another tidbit to keep in mind is that only switches support full duplex. If
>you use a hub (not sure how many still exist), you are forced to use half
>duplex.
>
>Sometimes, when using a switch, your NIC may autodetect a 100Mb half duplex
>connection. This is a huge bottleneck, as the NIC can only send or receive,
>not both, at the same time. Therefore, always, always check your link. Don't
>trust that autonegotiate will do what is best. You'll always want a full
>duplex connection over a half duplex connection any day.
>[...]
>
-- http://www.freelists.org/webpage/oracle-lReceived on Sat Dec 24 2005 - 01:27:00 CST
![]() |
![]() |