Definition of Congestion
The spooks have been working overtime, the NSA spent $1.5 million dollars this week recovering from a computer failure starting Monday. It was restored on Thursday, just about the time people started reporting strange problems on the Internet. Concidence? But on a more serious note. Alan used the word "congestion" in one of his messages. Congestion doesn't seem to have a very strict definition. It covers everything from a multi-week service interruption, to a dropped packet which triggers slow-start. Should the Internet follow the nuclear and power industry and come up with a set of standarized terms for different degrees of events: blackout, brownout, flicker, etc. Or follow the telecommunication industry which uses a single word, congestion, for most problems.
Should the Internet follow the nuclear and power industry and come up with a set of standarized terms for different degrees of events: blackout, brownout, flicker, etc. Or follow the telecommunication industry which uses a single word, congestion, for most problems.
IOPS attempted to do something similiar, at least for IOPS members. However, the document is publicly available, and goes on to provide methodologies to define and measure latency loss, so that when one ISP talks to another, they are talking close to the same language. The doc can be found at: http://www.iops.org/Documents/PacketLoss.html /Sean
On Sun, Jan 30, 2000 at 12:35:18AM -0800, Sean Donelan wrote:
Should the Internet follow the nuclear and power industry and come up with a set of standarized terms for different degrees of events: blackout, brownout, flicker, etc. Or follow the telecommunication industry which uses a single word, congestion, for most problems.
<sigh> it took this long for someone to ask that question in a public place? I think we should go further than that, and develop an internet standard for noc-noc communications. at least some sort of bcp, so that those who conform can know what to expect when calling a peer (in the literal, not necessarily BGP sense) noc who also claims to conform to such an ad-hoc "standard". for instance, it would be nice to consistently be able to get a peer noc to open a ticket on reported troubles, get them to cooperate in tracking security (i.e. DoS attacks) problems, etc. in 30 years, the ability for computers to converse with eachother and negotiate reasonably has improved exponentially. there are still humans sitting behind them however, who still have problems with this after several thousand years of evolution. $0.02 +/- 0.02 Sam -- Sam Thomas Geek Mercenary
I think this is the real issue here http://www.cert.org/incident_notes/IN-99-07.html tribe and trinoo Sam Thomas wrote:
On Sun, Jan 30, 2000 at 12:35:18AM -0800, Sean Donelan wrote:
Should the Internet follow the nuclear and power industry and come up with a set of standarized terms for different degrees of events: blackout, brownout, flicker, etc. Or follow the telecommunication industry which uses a single word, congestion, for most problems.
<sigh> it took this long for someone to ask that question in a public place?
I think we should go further than that, and develop an internet standard for noc-noc communications. at least some sort of bcp, so that those who conform can know what to expect when calling a peer (in the literal, not necessarily BGP sense) noc who also claims to conform to such an ad-hoc "standard". for instance, it would be nice to consistently be able to get a peer noc to open a ticket on reported troubles, get them to cooperate in tracking security (i.e. DoS attacks) problems, etc. in 30 years, the ability for computers to converse with eachother and negotiate reasonably has improved exponentially. there are still humans sitting behind them however, who still have problems with this after several thousand years of evolution.
$0.02 +/- 0.02
Sam
-- Sam Thomas Geek Mercenary
-- Thank you; |--------------------------------------------| | Thinking is a learned process so is UNIX | |--------------------------------------------| Henry R. Linneweh
Attempts to "standardize" inter-NOC communication happened a couple of times in the early 90's in the IETF, but it got too ugly, with too many barriers (basically people not wanting to expose dirty laundry). I fear that attempts to do so now when the providers are *really* competing with each other (as opposed to back in the old daze when we were all friends :-) will suffer the same fate, other than on a pairwise basis between NOCs that cooperate well with each other anyway. dave At 01:19 PM 1/30/00 +0000, Sam Thomas wrote: On Sun, Jan 30, 2000 at 12:35:18AM -0800, Sean Donelan wrote:
Should the Internet follow the nuclear and power industry and come up with a set of standarized terms for different degrees of events: blackout,
brownout,
flicker, etc. Or follow the telecommunication industry which uses a single word, congestion, for most problems.
<sigh> it took this long for someone to ask that question in a public place? I think we should go further than that, and develop an internet standard for noc-noc communications. at least some sort of bcp, so that those who conform can know what to expect when calling a peer (in the literal, not necessarily BGP sense) noc who also claims to conform to such an ad-hoc "standard". for instance, it would be nice to consistently be able to get a peer noc to open a ticket on reported troubles, get them to cooperate in tracking security (i.e. DoS attacks) problems, etc. in 30 years, the ability for computers to converse with eachother and negotiate reasonably has improved exponentially. there are still humans sitting behind them however, who still have problems with this after several thousand years of evolution. $0.02 +/- 0.02 Sam -- Sam Thomas Geek Mercenary
On 01/31/00, dave o'leary <doleary@juniper.net> wrote:
Attempts to "standardize" inter-NOC communication happened a couple of times in the early 90's in the IETF, but it got too ugly, with too many barriers (basically people not wanting to expose dirty laundry). I fear that attempts to do so now when the providers are *really* competing with each other (as opposed to back in the old daze when we were all friends :-) will suffer the same fate, other than on a pairwise basis between NOCs that cooperate well with each other anyway.
Those of us who are willing to communicate with each other should go ahead and write up a BCP, because the folks who're scared to admit that they're fallible won't follow it anyway. ---------========== J.D. Falk <jdfalk@cybernothing.org> =========--------- | "One computer is a problem. A computer network is a large problem. | | The internet is the world's largest problem." | | -- Douglas Warren | ----========== http://www.cybernothing.org/jdfalk/home.html ==========----
> Those of us who are willing to communicate with each other > should go ahead and write up a BCP, because the folks who're > scared to admit that they're fallible won't follow it anyway. I agree entirely. I can't stand it when people are too sheepish to publish Best Common Practices just because there are a few companies that don't feel like putting in the effort to do a halfway decent job. God knows we wouldn't want anyone to feel left out, even if they do insist on running open relays, or redistributing their IGP or whatever. Bleah. -Bill
At 03:30 PM 01/31/2000 -0800, dave o'leary wrote:
Attempts to "standardize" inter-NOC communication happened a couple of times in the early 90's in the IETF, but it got too ugly, with too many barriers (basically people not wanting to expose dirty laundry). I fear that attempts to do so now when the providers are *really* competing with each other (as opposed to back in the old daze when we were all friends :-) will suffer the same fate, other than on a pairwise basis between NOCs that cooperate well with each other anyway.
dave
I have to chuckle reading this thread. We can't even get our NOCs to talk to each other, and Marketing people think we can do Inter-Provider QOS?? Hopefully I will have retired at an early age before we burn that bridge :) -Steve
We can't even get our NOCs to talk to each other, and Marketing people think we can do Inter-Provider QOS??
Sure we can, its just not what the vendors think of QoS, and it might not be a very good sort of QoS. :)) Deepak Jain AiNET
participants (9)
-
Bill Woodcock -
dave o'leary -
Deepak Jain -
Henry R. Linneweh -
J.D. Falk -
Sam Thomas -
Sean Butler -
Sean Donelan -
Steve Meuse