Date: Thu, 03 Jun 1999 23:33:52 -0700 From: bmah@CA.Sandia.GOV (Bruce A. Mah) To: Arabian <Arabian@DAL.NET> Cc: Joel.Clark@scott.af.mil, bmah@california.sandia.gov, wes@softweyr.com, net@FreeBSD.ORG Subject: Re: Routers and such Message-ID: <199906040633.XAA20315@stennis.ca.sandia.gov> In-Reply-To: Your message of "Thu, 03 Jun 1999 19:30:58 %2B0300." <3.0.6.32.19990603193058.007c1b10@qatar.net.qa>
next in thread | previous in thread | raw e-mail | index | archive | help
If memory serves me right, Arabian wrote: > My server on ISP powered by FreeBSD also my own server is P II 333 with 128 > MB SDRAM my NIC is Intel EtherExpress Pro 10/100B Ethernet FreeBSD > 3.2-STABLE, the router where my server connected to on the ISP is also P II > 333 powered by FreeBSD. > > Could that router being problem and causing any latency for my serverand I > should ask them to connect it to the Core directly ? I'll try to answer this, but I have to admit that I'm not completely understanding the situation. I think you have a server of some sort running FreeBSD. It sounds like it's collocated at your ISP, and you have another FreeBSD machine acting as a router between your server and some subnet that's run by the ISP. You want to know if having that router in the middle is adding excessive latency, and if the server should consequently be attached directly to the ISP's subnet. Is that right? Without some more details, it's difficult to answer your question. For example, what kind of services are you providing? Where are the clients? Are the networks 10Mbps or 100Mbps? Is the router doing anything other than routing packets? What other hosts are the networks on each side of your router? Why is the router there in the first place? The big questions: Do you think you are actually experiencing any performance problems, and if so, could you quantify this? If I understand your environment correctly, I think that having the router in between shouldn't be a problem at all for you. For sake of argument, if your server is serving Web pages to random clients on the Internet, it's a fair guess that there's going to be a lot more latency between the clients and your ISP than there's going to be through your router. (For something to try, login to your server and ping random hosts on the Internet, observing the round-trip time statistics. Then login to your router and do the same. Compare the round-trip times in both cases.) > What is the different between half and full duplex lans ? In the context of full-duplex vs. half-duplex 10baseT or 100baseTX Ethernet: In half-duplex, only one system on a cable can be transmitting a packet at a time. In the common case of a host attached to a hub, this means that the host can only either be transmitting a packet or receiving a packet but not both. This situation is analogous to the original thick and thin Ethernets, which were long coaxial cables. Access to transmit on the cables was (is) shared between all the attached hosts. An important part of the Ethernet standard is the algorithm by which hosts contend for access to this shared network. With Ethernet switches, you have the option of full-duplex Ethernet transmission. This basically means that a host can be both transmitting a packet and receiving a packet on the cable at the same time. Hope this helps... Bruce. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199906040633.XAA20315>