Quote:
Originally Posted by bomber_g
You personal experience does not really give a good overall view of what to expect when you are trying to design resilient service for someone else.
|
Except it does when that's what I do (did) for a living. Sure core networks are physically more resiliant, but most outages are caused by operator error or deliberate action (e.g. maintenance) and complete failures of resilient hardware is indeed very rare. However, operator issues don't affect the access network nearly as much, people don't go around upgrading or re-routing the access network on a daily/weekly basis, and where they do cock up, it affects only one customer, not a hundred thousand.
While there may be more hardware faults on the access network, they usually affect no more than one individual or small group of people. Even a minor core network fault affects thousands to millions of users. If one core network fault occurs per year, affecting 1 million users, and access network issues happen a thousand times per day, that's
still three times more end user outages caused by a single core fault than 365,000 access network faults.
Admittedly I didn't work for BT, but since my consumer experience of BT and VM both reflect my work experience, I'm not convinced. I've seen five 1m+ outages in the last 6 months, and with only 2000 Openreach engineers in the country, each one would have had to be fixing 14 per day, every day, seven days a week for there to have been more access network outages.