On Wed, 24 Jan 2001, Henry Yen wrote: > On Wed, Jan 24, 2001 at 10:17:19AM -0500, Richard A. Steenbergen wrote: > > Super-redundant DNS hosting would make a good product. Akamai maybe? > > seems to me if a company bought a dedicated server in a couple of > dozen of the largest hosting sites, that might be enough. > you could run dns, remote network monitoring, traffic stats, whatever. > > does anyone out there know of such a service? That's what Nominum's GNS service is. It's hosting some cc-tlds and whatnot at the moment, although they haven't started selling it to smaller customers yet. http://www.nominum.com/GNS/index.html No details on the site, but David Conrad or Paul Vixie or I can tell you more about it. Basically it's a global network of redundant pairs of heterogenous nameservers (Solaris on Netra UltraSparc and NetBSD on IBM Netfinity x86) with IP Anycast so you always get the nearest pair. For example: MAE-West Looking Glass Results Tracing the route to gns1.nominum.net (198.133.199.1) 1 mae-west.above.net (198.32.136.31) 0 msec 0 msec 4 msec 2 core1-core2-oc3-2.sjc6.above.net (209.133.31.177) [AS 6461] 0 msec 4 3 pao1-sjc6-oc12.pao1.above.net (207.126.96.54) [AS 6461] 4 msec 0 msec 4 64.124.11.252.nominum.com (64.124.11.252) [AS 6461] 0 msec 0 msec 4 5 gns1.nominum.net (198.133.199.1) [AS 17204] 4 msec 0 msec 0 msec MAE-East Looking Glass Results Tracing the route to gns1.nominum.net (198.133.199.1) 1 iad1-core4-pos1-0.atlas.icix.net (165.117.52.186) 4 msec 0 msec 0 msec 2 dca6-core2-pos1-3.atlas.icix.net (165.117.63.5) 0 msec 4 msec 0 msec 3 dca6-core3-pos7-0.atlas.icix.net (165.117.48.110) 0 msec 4 msec 0 msec 4 above-digex-oc3.iad.above.net (216.200.127.113) 0 msec 0 msec 4 msec 5 core5-core3-oc48.iad1.above.net (208.185.0.150) [AS 6461] 0 msec 4 6 core2-iad1-oc48.iad5.above.net (216.200.127.14) [AS 6461] 0 msec 4 7 above-gw.nominum.com (216.200.105.54) [AS 6461] 4 msec 4 msec 0 msec 8 gns1.nominum.net (198.133.199.1) [AS 17204] 4 msec 4 msec 0 msec Both of those traces are running through AboveNet, but locations are BGP transit multi-homed and peer with a lot of providers, and since it's stateless UDP transactions, all the fail-over is transparent anyway. -Bill