Re: [Raw] FW: New Version Notification for draft-pthubert-raw-architecture-00.txt

Hi Pascal,

Reliability and Availability are unfortunately controversial terms as different industries, or even sub-groups of industries have defined them independently of each other, hence, very much unfortunately differently. A lot of discussion is now happening on it in multiple fora with the convergence of Operations Technology, Information Technology, and Telecommunications networking. I would not try to redefine them in RAW. We may consider taking one of the existing or converged definitions.

When it comes to packet loss, I have the same information as you: there is no tolerance of n consecutive packets in industrial automation for the reasons you described. The numbers I heard is no 2 or 3 consecutive packets are allowed to be lost.
This is also captured in the DetNet use cases:
https://tools.ietf.org/html/rfc8578#section-7.4
In DetNet, we did not try to convert this requirement to X nines, but introduced new parameter for it:
https://tools.ietf.org/html/draft-ietf-detnet-flow-information-model-07#section-5.9.5

As it was also pointed out on the DetNet list, it actually often means zero packet loss requirement:
https://mailarchive.ietf.org/arch/msg/detnet/nkoqM-cBlmIhBoyvs4u6nv-q-t0/

So, we need service protection to meet this requirement as described in DetNet Architecture: https://www.rfc-editor.org/rfc/rfc8655.html#name-service-protection.

Stay safe,
Janos

-----Original Message-----
From: RAW <raw-bounces@ietf.org> On Behalf Of Pascal Thubert (pthubert)
Sent: Thursday, April 2, 2020 4:04 PM
To: Rex Buddenberg <buddenbergr@gmail.com>; raw@ietf.org
Cc: Fabrice Theoleyre <theoleyre@unistra.fr>; CARLOS JESUS BERNARDOS CANO <cjbc@it.uc3m.es>
Subject: Re: [Raw] FW: New Version Notification for draft-pthubert-raw-architecture-00.txt

Dear RAWers,

The Industrial people I've been discussing with tell me that they want 5 nines of reliability (PDR=99.999%) because if the faults are independent, then there are 10 nines that we have 2 faults in a row, and 15 nines that you get three.

What they really want is to avoid faults in a row. Because after 4 packet losses, the control loop is unstable and the production line is emergency halted. Once a year you get a bad look. Twice and your equipment is on the curb.

You see that the nines of reliability actually translate in an unscheduled loss of availability that is less than 4 loops. Because we cannot repair that fast, we need redundancy is a requirement.

In the architecture I used the following definition:

   Reliability:  Reliability is a measure of the probability that an
      item will perform its intended function for a specified interval
      under stated conditions.  For RAW, the service that is expected is
      delivery within a bounded latency and a failure is when the packet
      is either lost or delivered too late.  RAW expresses reliability
      in terms of Mean Time Between Failure (MTBF) and Maximum
      Consecutive Failures (MCF).

   Availability:  Availability is a measure of the relative amount of
      time where a path operates in stated condition, in other words
      (uptime)/(uptime+downtime).  Because a serial wireless path may
      not be good enough to provide the required availability, and even
      2 parallel paths may not be over a longer period of time, the RAW
      availability implies a path that is a lot more complex than what
      DetNet typically envisages (a Track).

(feedback welcome!)

Certainly the ratio of usable time is of interest. If it's scheduled and you can coordinate other maintenance activities then it can last longer. But the losses in a row are the real criterion on which automation operates; the reliable cross-over that Rex tells us about must happen within that time.

Keep safe,

Pascal

-- 
RAW mailing list
RAW@ietf.org
https://www.ietf.org/mailman/listinfo/raw