> On Fri May 30 2008 13:37:12 Dotan Cohen wrote: >> > It's your ISP's job to find out why the connection is timing >> > out. If they can't or won't, trying to explain "mtr" to them >> > isn't going to help. (BTW, your "mtr" was bad but not awful >> > because of the "2.3%" line.) >> >> Could you elaborate on that? I've stfw but I'm still not quite >> proficient with mtr or computers in general. > > If an mtr or traceroute or tracepath, after a sufficient number of > iterations, looks like this: > > HOP1 20% packet loss > HOP2 30% packet loss > HOP3 2% packet loss > ... > > - it means that 98-100% of packets are getting to HOP3 and back, > which means that the packet loss at the earlier hops is probably > due to high CPU loads or ICMP rate limiting rather than fundamental > loss of packets in transit. (But there could be more esoteric > causes such as errors in diverse return routes.) > >> > If you really have to use "mtr", first of all compare a regular >> > "mtr" with a "mtr -i 10" to see if DOS attacks (and "mtr") are >> > being filtered. >> >> I suppose that I don't _have_ to use mtr but it is the tool that I've >> heard of. I changed the interval time to 10 seconds like you mention, >> but I don't see where that helps. > > If you're sending N test packets per second you may be running afoul > of an ISP's anti-DOS-attack filters. By comparing those results with > results obtained at N/10 test packets per second, you can determine > whether this is a likely explanation. > >> Is there a resource that you could recommend that I read to learn to >> be more proficient with mtr? Written for silly end-users like me? > > Sorry, I don't know of one. "mtr" is a tool. A rough analogy would > be me asking for instructions on how to use a screwdriver when my > car is broken. I would need to understand how the car's systems > work in order to understand how to find out what is wrong and how to > fix it. > > There are lots of good introductory networking courses available. > I recommend Cisco's CCNA but you might find others more to your taste. > I won't pretend that CCNA (or even CCNP) will explain the full > implications of "mtr" output, but they'll give you enough background > that you will then be able to understand the implications. > > --Mike Bird
There is a graphical traceroute program available at changeip.com. In general though they all work roughly the same way:a packet is sent from your source to a destination with the Time To Live (TTL) field set=1 and also sets a timer. The first router decrements the TTL and since it is now '0' returns an ICMP packet with the router as source. When the originator gets it he/she knows the identity of the first router and how long it took. The process repeats until the entire path is traced and the statistics of each of the hops can be measured. HTH Larry > > > -- > To UNSUBSCRIBE, email to [EMAIL PROTECTED] > with a subject of "unsubscribe". Trouble? Contact > [EMAIL PROTECTED] > > -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
