| gabriel rosenkoetter on Tue, 22 Oct 2002 13:10:06 -0400 |
|
On Mon, Oct 21, 2002 at 10:58:27PM -0400, William H. Magill wrote: > Bill's traceroute points out a problem that I see more and more > frequently all around the net these days... INCREDIBLY HUGE hop-counts. > Hop-counts have been steadily increasing around the net for years, so > this kind of hop-count is actually pretty "normal" these days. But > while hop-counts in and of themselves are not a significant issue, the > fact is that they put that many more places for congestion to occur > between you and your destination... "Hop-count" is essentially meaningless. All it means is that there's more hardware that decrements TTL between you and the remote host; it doesn't necessarily mean that that transit will be slower. You see increased hop counts because Tier 1 providers are using more and more edge routing hardware that decrements the TTL. That all *should* be connected with fibre going way faster than you and the whole neighborhood could possibly saturate with a cable modem (the constriction is *far* more likely to be where all those cable lines come together to get back to the provider). That said, there've been some telco problems lately (heard of Worldcom? Not to mention the Qwest fiasco several months ago in Europe). And, in any case, Bill's hops to Paul's DNSes are pretty irrelevant; the chances of Bill having to do a DNS lookup against them in the normal course of operations is extremely slim... that's what the D in DNS tries to avoid. > and the current IP default is 30 hops! (It used to be 10.) Huh? You mean traceroute(8)'s default? What bearing has that on networking? -- gabriel rosenkoetter gr@eclipsed.net Attachment:
pgp19MsoO84hN.pgp
|
|