Re: [arch-d] possible new IAB programme on Internet resilience

On 23-Dec-19 22:12, Yaakov Stein wrote:
> Resilience is widely used in transport networks as a synonym for Automatic Protection Switching,
> which typically replaces a failed path in less than 50 milliseconds.
> So, while I agree that resilience is the ability to respond to an actual failure, it is not hypothetical or necessarily of long time duration.
> 
> Robustness is used in system design and computer science to mean the ability to avoid failure despite disturbances or unexpected inputs.
> Failure is defined (e.g., G.805) as the inability to provide the service or to meet the user's expectations.
> Note that unlike resilience, robustness is always with respect to some disturbance type, and that a system can be robust to X but not to Y.
> For example, a network designed to be robust with respect to link failures will contain a lot of redundant links,
> but then it will not be robust with respect to increasing traffic levels (and vice versa).

Yes. I don't think there is a sharp line between resilience and robustness. Also, despite the IETF's traditional narrow scope of focus on the wire protocol, you can't build a robust or resilient network without considering more than just the wire protocol. We might have to step outside our comfort zone here. For example, there is no such thing as a fatal error in a resilient system. All exceptions must be caught. That's a stronger statement than saying that a protocol's state machine must contain no undefined arcs.

   Brian

> 
> Y(J)S
> 
> -----Original Message-----
> From: Architecture-discuss [mailto:architecture-discuss-bounces@ietf.org] On Behalf Of Stephane Bortzmeyer
> Sent: 23/12/2019 10:00
> To: Stephen Farrell <stephen.farrell@cs.tcd.ie>
> Cc: architecture-discuss@iab.org
> Subject: Re: [arch-d] possible new IAB programme on Internet resilience
> 
> On Sat, Dec 21, 2019 at 01:23:18AM +0000,  Stephen Farrell <stephen.farrell@cs.tcd.ie> wrote  a message of 271 lines which said:
> 
>>> Indeed, the current project of charter seems to be more about 
>>> robustness than about resilience.
>>
>> Sorry - I'm not really following any of the above comments.  Can one 
>> of you try explain some more?
> 
> Typically, robustness is the ability to withstand shocks without cracking, and resilience is the ability, once you have cracked, to restart afterwards. I suggest that the word "robustness" is a better fit for what we are trying to achieve (an Internet that will keep going, may be in degraded mode, when problems/attacks occur). "Resilience" would be adapted for a (very hypothetical) case where everything was down and we try to resume activity. (Thought
> experiment: all the DNS root name servers were destroyed and we must agree on a new set, without beging able to communicate together, except by IP addresses.)
> 
> _______________________________________________
> Architecture-discuss mailing list
> Architecture-discuss@ietf.org
> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Farchitecture-discuss&amp;data=01%7C01%7Cyaakov_s%40rad.com%7Cf65119756ea34c3adefc08d7877e387a%7Cf9047108cc2c4e4897a343fad1b3bf9d%7C1&amp;sdata=%2F60sYdn%2Ba5FZ%2FCO5tyBP1quAAlLF2t7tanBy4Al41sU%3D&amp;reserved=0
> 
> _______________________________________________
> Architecture-discuss mailing list
> Architecture-discuss@ietf.org
> https://www.ietf.org/mailman/listinfo/architecture-discuss
> 