What do y'all think of pathchar *as it is now*? How reliable is it for determining actual _available bandwidth_? I thought according to VJ, CAIDA, Cisco and even the NSF it was "way-alpha"? Steve Blair wrote:
well, one could get Van Jacobson's pathchar, and learn the true capacity that way. I could care less what *theoretical bandwidth* is available, when customers complain. I want to know what bandwidth *is available*, and pathchar goes a very long way towards that end.
link labelling could be arguably silly, if you're basing your determination on a competetitor's labels, you could get some unusual surprises IMHO...
-- steve c blair tivoli systems inc sblair@dev.tivoli.com "Why can't we blast them onto someone else's property?"
'Vadim Antonov writes...' **> **> There's no use in promoting corporate paranoia at the expense of **> engineering cooperation. It is like butcheing the hen which lays **> the golden eggs. Knowing link capacity was useful (while it lasted) **> to get the idea of what is more likely to be dropping packets **> on the floor when customers complained.
On Thu, 13 Nov 1997, Ex-Support Wench wrote: I thought it has a lot of great potential usage; however, I was dismayed at how much traffic it generates. As far as way-alpha goes, it still requires a bit of tuning. I tried the first edition, and it constantly reported my 10 Mbit ethernet as an 83 Mbit cable. While I like the idea that my network equipment might be able to suck the bits right out of the machine, making them move faster, I noticed it needed some accuracy adjustments. (Note: The shortly-thereafter-released second version did fix that and brought the bandwidths down.) /cah ==>What do y'all think of pathchar *as it is now*? How reliable is it for ==>determining actual _available bandwidth_? I thought according to VJ, ==>CAIDA, Cisco and even the NSF it was "way-alpha"? ==>
Hello alex, On Thu, 13 Nov 1997, Ex-Support Wench wrote:
What do y'all think of pathchar *as it is now*? How reliable is it for determining actual _available bandwidth_? I thought according to VJ, CAIDA, Cisco and even the NSF it was "way-alpha"?
From the testing I've done the results are a reasonable approx. of the -available- bandwitdh. I have also seen some very outragious output out of the utility (solaris). The linux version isn't able to figure out the MTU so you have to specify it on the command line, when traversing a slower link the out put from the far side can be -very- un-reliable. Hth, JimL
Steve Blair wrote:
well, one could get Van Jacobson's pathchar, and learn the true capacity that way. I could care less what *theoretical bandwidth* is available, when customers complain. I want to know what bandwidth *is available*, and pathchar goes a very long way towards that end.
link labelling could be arguably silly, if you're basing your determination on a competetitor's labels, you could get some unusual surprises IMHO...
-- steve c blair tivoli systems inc sblair@dev.tivoli.com "Why can't we blast them onto someone else's property?"
'Vadim Antonov writes...' **> **> There's no use in promoting corporate paranoia at the expense of **> engineering cooperation. It is like butcheing the hen which lays **> the golden eggs. Knowing link capacity was useful (while it lasted) **> to get the idea of what is more likely to be dropping packets **> on the floor when customers complained.
From my experience, I have yet to see it give realistic numbers on OC3 and OC12 circuits.
Ex-Support Wench writes:
What do y'all think of pathchar *as it is now*? How reliable is it for determining actual _available bandwidth_? I thought according to VJ, CAIDA, Cisco and even the NSF it was "way-alpha"?
Steve Blair wrote:
well, one could get Van Jacobson's pathchar, and learn the true capacity that way. I could care less what *theoretical bandwidth* is available, when customers complain. I want to know what bandwidth *is available*, and pathchar goes a very long way towards that end.
participants (5)
-
Craig A. Huegen -
Ex-Support Wench -
Henry Kilmer -
Network Operations Center -
Randy Bush