06-03-2002 03:44 AM - edited 03-01-2019 09:59 PM
How can I verify the speed between two locations connected via a DS3?
06-03-2002 07:18 AM
I'm not sure what you mean by "verify the speed." Through a router? If it is the DS3 itself you are concerned about, you needn't worry. A DS3 is a DS3 is a DS3 (44.736 Mbps). If you have particular set of hosts and you want to test TCP/IP throughput between them, I would recommend that you look into TTCP (test TCP) or something similar. I will caution you on a couple of things: First of all, you need to consider your "bandwidth * delay product," which is a function of bandwidth and delay. A 45 Mbps, if I'm doing my math right, your round-trip delay needs to be limitted to about 12 msec or performance will start to drop off as a result of an exhausted tcp window (this, of course, assumes you are trying to achieve 45 Mbps on a single TCP socket - not something that is commonly done). Also, many host processors aren't capable of sustaining traffic at 45 Mbps.
06-03-2002 11:09 AM
I am looking to ensure that the DS3 and the routers on each end are operating at full efficiency.
06-03-2002 03:18 PM
I'm afraid it isn't that straight-forward. You could have routers A & B separated by, say, North America. Those two routers could operate at "full efficiency" (assuming the circuit is operating within specs) if you had a couple of hundred hosts all doing various udp, tcp, ftp, etc transactions. If you had two hosts trying to consume that entire pipe with a single tcp socket over that distance, you'd have some serious problems.
Characterizing traffic flows and network performance is as much an art as a science. There isn't really any one single magic show command that I'm aware of that will satisfy your curiosity. You can, of course, do basic 'sh int' several times to look at the load as a weighted 5 minute average (or run MRTG to get nice looking graphics). This will also alert you to any serious layer 2 errors that might be occurring on the circuit (SES, etc). You can also look at misc. stuff such as queuing, processor usage, and memory usage to ensure that the router isn't being killed. But none of these can allow you to sleep at night knowing that everything is at full efficiency. You have to get down and sniff individual flows, test throughput between hosts using ttcp and others, test response times and so on to get that level of comfort.
Having said all of that, if you have reason to suspect something fishy, I would recommend that you check your basic config. I once had an out-of-town client who insisted for several months that the SONET span between two DS3-equipped routers was magically providing timing. They even brought me out and put test sets on the circuit to prove it was running "error-free." Then I showed them the thousands of clock slips and unavailable seconds they were counting up along side that "zero bit errors" display. So we set one of the routers to generate clock and "performance," as measured by the number of user complaints, went up by orders of magnitude.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide