Hi all, Today was the first day we operated on the Linux Ethernet Bridge. It didnt lock up or lose a packet once! I hope you are proud Lennert. Your software did its job better than i expected. Thank-you again.
Now to spend the $8000 Australian to buy the new gigabit fibre gear to fix our fibre backbone. When I first started at my workplace (2 yrs ago), I was told that the entire backbone was 100MBps. So i belevied the Admin since he put it in, he has since moved on to better things. I did some checking today and found that the entire backbone is 10MBps, useless. We have 100MBps switches running off 10MBps fibre. No wonder we are having problems. I have ordered in the new gear for the switches etc, and we also have been given $50,000 to replace the current microwave link, which is now illegal by regulations set by the Communications Authority here in Australia. The new link will be 100MBps Full Duplex. Our first site has 100 computers on it (the server site) and the other site has only 50 machines. So the 100MBps will be plenty, as long as there is a bridge in front of it. The bridge will be the Linux Ethernet Bridge. I am willing to write a testimonial once we get all this gear in, about 2 months away. Thanks again, Mathew ----- Original Message ----- From: "Lennert Buytenhek" <[EMAIL PROTECTED]> To: "Mathew McKernan" <[EMAIL PROTECTED]> Cc: <[EMAIL PROTECTED]> Sent: Friday, December 14, 2001 1:30 AM Subject: Re: [Bridge] Buffering of packets > > On Fri, Dec 14, 2001 at 12:05:20AM +1100, Mathew McKernan wrote: > > > > I'm not exactly sure what you mean with your question. If you are asking > > > whether linux ever drops packets, I can assure you that it will. If you > > > are bridging from 100Mbit to 10Mbit, it takes ten times longer to transmit > > > the data than to receive it, so if insist that no packets be dropped, even > > > if you put the 100Mbit under load for a day, you will need 10 days worth > > > of 10Mbit buffering capacity in the bridge, which is about a terabyte. > > > > I know we dont run our LAN that hard :P, but i see what your saying. > > One thing that might make a difference if you have wide variation in > link utilisation is altering the transmit queue length on the slow > interface, this effectively increases the amount of data that will be > buffered for sending (the default is 100 packets). This _will_ increase > forwarding latency under high load, though. It won't help the case > where you have constant high load. > > ip link set eth0 txqueuelen 200 > > > cheers, > Lennert > _______________________________________________ > Bridge mailing list > [EMAIL PROTECTED] > http://www.math.leidenuniv.nl/mailman/listinfo/bridge _______________________________________________ Bridge mailing list [EMAIL PROTECTED] http://www.math.leidenuniv.nl/mailman/listinfo/bridge
