At 4:42 PM 3/30/95, Hans-Werner Braun wrote:
I wonder by how much the problem could be reduced as such, if service providers (including campus service providers) would have accessible servers about network status information, being kept current by some local NOC.
At 4:56 PM 3/30/95, Chris Dorsey (510)422-4474 wrote:
.... Of course there will always be some NSP's that would never implement such a server for business purposes.
H-W, Chris, et al; Some service providers already post system-wide planned outages and even system-wide unplanned outages (like NSFnet backbone outages) on public mailing lists. We don't need another protocol to support this -- it would be a simple matter for everyone to gateway system-wide trouble ticket reports to a Web page. As Chris points out, even if today we could do this, in future it will become more difficult as competitive pressures mount. In the past I have managed NOCs where there were pressures on the NOC to sugar-coat the public trouble tickets. A certain amount of this is appropriate, since it is possible for NOC controllers, network engineers and other technical support people to become frustrated with their peers and I have some vivid memories of particular frustrations coming out in system-wide trouble tickets. It doesn't help either party for shouting and name-calling to show up in trouble tickets, but it can easily happen, as we all know. So when the pressure inevitably mounts on everyone to treat their system-wide trouble tickets like press releases, the information content that we seek will tend toward zero. Therefore I feel that if such a public system were created it would inevitably devolve to minimize useful information, such as who is really screwing up or where the difference of opinion actually lies. We need a new pressure point, like traceroute became for routing or throughput became for router benchmarks. If you all kept incident notes and someone sent out a survey every quarter, would you be interested in a Consumer Reports style of NOC performance metric? :-) This might be worth thinking about in the IP Perf Metric BOF next week. I just don't see any other pressure point. There has to be an outside evaluation tool and a general understanding of how to interpret it. --Kent