All networks are vulnerable to Denial of Service attacks and flash crowds. Broadcasting & Cable investigates the problems with the telephone and SMS voting with the Fox television show American Idol. http://www.broadcastingcable.com/article/CA417981 The difference is when people get a "busy" signal on the phone network, they don't think phone network is collapsing.
On 17-mei-04, at 4:53, Sean Donelan wrote:
All networks are vulnerable to Denial of Service attacks and flash crowds. Broadcasting & Cable investigates the problems with the telephone and SMS voting with the Fox television show American Idol.
This is _very_ old news. Something like this happened in Holland ages ago, and it was even named the "Henny Huisman effect" after the show's host. The phone network is very fragile. Just about anything unusual causes massive congestion. For instance, the 9/11 problems in New York disrupted long distance service throughout the entire US and even international service in some unrelated parts of the world. But even more mundane problems such as last summer's power outage have similar effects.
The difference is when people get a "busy" signal on the phone network, they don't think phone network is collapsing.
If they knew the difference between a busy signal and a congestion signal they probably would... BTW, what do you think happens when everyone starts running the water tap at the same time?
On Mon, 17 May 2004 10:32:32 +0200 Iljitsch van Beijnum <iljitsch@muada.com> wrote: | If they knew the difference between a busy signal and a congestion | signal they probably would... Er, no. Congestion signal normally means that there are no circuits and the phone network has handled that situation without any issues. But that's not the primary threat that the switches have to handle When the call demand far exceeds the number of circuits/operators available, repeated busy or congestion tones will cause callers to make repeat attempts. Local (originating) switches handle this just fine, and then send forward a C7 call set-up request to the switch that handles inbound for that number range (I guess, that's their equivalent of an "MX" host). And that's where things go wrong. Digital circuit-switches such as AXE10, DMS100/250 etc are far more vulnerable to high levels of call-set-up traffic, which would cause their processors to be overloaded. Again, the IP analogy is obvious. Multiple-repeat-attempts at call setup to the same number (and same destination switch) from numerous originating switches, cause the processor at the destination switch to be overloaded and to crash. That's doesn't result in busy or congestion signals - that results in NO signals (not even dialtone in that exchange's local area). The telcos' priority therefore is to block the call-setup-attempts at the edges of their network (i.e. originating /early transit switches). This is known as "call-gapping" and is not without some controversy. -- Richard Cox
Richard Cox wrote:
This is known as "call-gapping" and is not without some controversy.
Richard doesn't say - cause he's too polite - is that in the UK you can *buy* this service as a customer. Oh, I only want 1 in 20 calls to arrive please... This has started to die as more and more large call terminators (game shows, charities etc.) make money out of interconnect and non-geographic termination revenue. Now the objective is to terminate every call and keep the cumb pleb on hold as long as possible. Strange how I very rarely call a *sales* number now that is neither freephone or real geographic. I know, in the UK at least, that if the company has an "0870" (Netional Call Rate for non UK folks) sales number, then it is not in their interest to get me off hold quickly. The reason I say that is, historically at least, I recall that the provisioning of the network is different in different markets because of the economies of caller-pays vs. called-party-pays. Peter
On Mon, 17 May 2004, Iljitsch van Beijnum wrote: : > The difference is when people get a "busy" signal on the phone : > network, they don't think phone network is collapsing. : : If they knew the difference between a busy signal and a congestion : signal they probably would... In the US, the busy signal and reorder signal use the same tone (combined 480+620 Hz), but the reorder signal sounds every half second instead of every second. Reorder is used in a failure situation that is not otherwise handled -- typically because of a dead trunk, completely full outbound trunks, or timeout during call setup. Most folks don't have any idea what the reorder tone means. It just sounds like a "fast busy signal" to them, and is often confusing. In real overloaded congestion situations, most local and long distance providers do use recordings with SIT tones that typically follow one of the wordings: All circuits are busy now. Please try your call again later. Your call did not go through. Please try your call again. Where they can be used, most telcos prefer these to a reorder tone, as the human voice makes the network seem less fragile to users. On a tangent, a morbidly amusing recording from the AT&T long distance network that I heard in June 1992 when trying to call a relative in Laguna after a large earthquake hit California: Due to a natural disaster in the area you are calling, your call did not go through. Will you please try your call again later. -- -- Todd Vierling <tv@duh.org> <tv@pobox.com>
On Mon, 17 May 2004 10:32:32 +0200, Iljitsch van Beijnum said:
This is _very_ old news. Something like this happened in Holland ages ago, and it was even named the "Henny Huisman effect" after the show's host.
Wouldn't the first "flash crowd" be all the people watching the Ed Sullivan show, who all hit the bathroom at the same time when a commercial came on? No wait, that would be a flush crowd..... ;)
On 17 May 2004, at 04:32, Iljitsch van Beijnum wrote:
On 17-mei-04, at 4:53, Sean Donelan wrote:
All networks are vulnerable to Denial of Service attacks and flash crowds. Broadcasting & Cable investigates the problems with the telephone and SMS voting with the Fox television show American Idol.
This is _very_ old news. Something like this happened in Holland ages ago, and it was even named the "Henny Huisman effect" after the show's host.
There was a similar thread on this list about cellular and international PSTN circuit congestion in New Zealand as the clock ticked over from 1999 to 2000, e.g.: http://www.merit.edu/mail.archives/nanog/1999-12/msg00390.html http://www.merit.edu/mail.archives/nanog/1999-12/msg00417.html Joe
participants (7)
-
Iljitsch van Beijnum
-
Joe Abley
-
Peter Galbavy
-
Richard Cox
-
Sean Donelan
-
Todd Vierling
-
Valdis.Kletnieks@vt.edu