title: Emergency Service in a Best-Effort World


by: Scott Bradner


For many years the phone system has been able to give special handling for selected phone calls.  This feature is designed to be used in times of emergency by medical services, some government officials, fire fighters, police, and some industry emergency response teams. A year or two ago some people started trying to figure out how to provide emergency services over the Internet.  The experiences in the aftermath of the events of September 11th made thinking about this both more and less important.


The current system in the U.S. is known as The Government Emergency Telecommunications Service (GETS).  This PIN-based system provides expedited handling of call requests but does not include preemption of calls already in progress.  It is provided by telephone companies under a government fee-for-service contract.


Clearly, with the movement towards converged networks, it makes sense to look at the impact of emergencies on vital Internet-based services.  But just understanding what vital services might encompass on the Internet is not easy.  In the phone world there is basically one service -- a fixed bandwidth voice call.  In the Internet there are hundreds of applications that might be important when responding to emergencies.  Dealing with each of the applications individually would be a daunting task, made all the harder by the people that keep creating new applications.


One proposal that is being discussed in the IETF, the ITU-T and ETSI is based on the International Emergency Preparedness Scheme (IEPS).   (www.iepscheme.net) As you can see if you take a look at the mailing list archives (reachable through their web page), this proposal has created some spirited discussions. 


Much of the discussion concerns the fundamental differences between the circuit-based, guaranteed quality, access controlled phone network and the packet-based, best-effort Internet and what did and did not happen on September 11th.  The Internet infrastructure did not collapse on September 11th but many web servers and some tail circuits were way overloaded.  This means that special traffic handling of emergency related traffic in ISP backbones may be much less important than ensuring priority access to network-based servers or tail circuits.


Much more work needs to be done to understand just what should be done in this area and just as important, what is not worth the effort to do.


disclaimer: Now who would claim that Harvard is not worth the effort?  Maybe MIT.  Anyway, the above observation is my own.