To:
Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>
cc:
randy@psg.com (Randy Bush), dnsop@cafax.se
From:
marka@isc.org
Date:
Thu, 08 Jul 1999 23:15:04 +1000
In-reply-to:
Your message of "Thu, 08 Jul 1999 19:52:00 +0200." <199907081052.TAA15394@necom830.hpcl.titech.ac.jp>
Sender:
owner-dnsop@cafax.se
Subject:
Re: Topological Motivation for draft-ohta-root-servers-01.txt?
> Randy; > > > >> We should be defining the siting of servers on technical > > >> grounds not politcal ones. > > > Do you also think so? > > > > yes. > > Then, according to Mark's example, if there is a root server in > France and there is no technical grounds to have one more in Monaco, > what will happen? Nothing. Will they get significantly better service? I doubt it unless the server in France is overloaded. > > Do you want to stop people in Monaco operate their own root > servers? I said there was no need, not that they can't do it or that I wish to prohibit them doing it. The siting of server should be based on need. I think Australia needs root servers based on excessive RTT to get to the current set of root servers. Only 30ms of the times below is related to my local link. After that all the links until the packet gets out out the country are 34M or better and reasonably unloaded. You will note Japan hosts the closest root server. A round-trip min/avg/max = 538.075/564.123/618.585 ms B round-trip min/avg/max = 395.34/414.388/511.844 ms C round-trip min/avg/max = 551.811/574.045/636.487 ms D round-trip min/avg/max = 545.681/564.311/622.53 ms E round-trip min/avg/max = 482.533/502.24/570.518 ms F round-trip min/avg/max = 481.896/495.207/528.304 ms G round-trip min/avg/max = 555.316/568.384/615.337 ms H round-trip min/avg/max = 460.838/491.312/568.309 ms I (no response) J round-trip min/avg/max = 530.988/560.705/637.963 ms K round-trip min/avg/max = 627.765/658.583/816.851 ms L round-trip min/avg/max = 391.142/412.013/480.743 ms M round-trip min/avg/max = 233.128/264.374/419.544 ms Another technical reason for adding an additional root should be to load share. And yet another to provide redunancy against single point failures. > > If you don't want it, what is the point of defining the siting? So that we get resonably uniform service levels wherever you are in the world. This is patently obvious that this is not the case at the moment. Below are the RTT as measured from California. The worst cases below are better than the best cases above. A round-trip min/avg/max = 64.002/64.386/64.73 ms B round-trip min/avg/max = 16.22/16.717/17.87 ms C round-trip min/avg/max = 90.174/110.793/233.235 ms D round-trip min/avg/max = 190.128/216.845/235.791 ms E round-trip min/avg/max = 7.522/8.081/9.097 ms F round-trip min/avg/max = 4.644/4.737/4.817 ms G round-trip min/avg/max = 179.227/394.412/691.453 ms H round-trip min/avg/max = 75.882/76.917/80.992 ms I round-trip min/avg/max = 162.517/166.611/170.374 ms J round-trip min/avg/max = 66.372/66.617/67.103 ms K round-trip min/avg/max = 150.063/151.216/152.626 ms L round-trip min/avg/max = 19.218/19.851/20.908 ms M round-trip min/avg/max = 159.566/170.415/262.884 ms We can make the technology find the close servers, but they need to be there in the first place. We need to define what is acceptable close. It does not matter if close is just across a border. Mark > > If you want it, how can you enforce it? What if, Monaco government > legislates that Internet users and providers in Monaco must use > Monaco local root servers? > > Masataka Ohta > -- Mark Andrews, Internet Software Consortium 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: marka@isc.org