Comments on: Too Connected, Or Not Connected Enough?

By: Daniel Tunkelang

Daniel Tunkelang — Fri, 24 Apr 2009 17:31:18 +0000

I’m sympathetic to planners who have to confront the enormous costs of replacing a legacy infrastructure like our power grid. But I worry when we make fresh decisions to build critical functions that depend entirely on external services with no backup plan if someone cuts a handful of wires. You’d think we’d have learned our lesson by now.

By: Andrew

Andrew — Fri, 24 Apr 2009 16:09:30 +0000

Daniel, to your point about our home town, don’t forget about this event (http://en.wikipedia.org/wiki/2003_North_America_blackout) that happened only 5 and a half years ago. Although it wasn’t triggered by sabotage, our infrastructure is in desperate need for additional resiliency and modernization.

By: Max L. Wilson

Max L. Wilson — Fri, 24 Apr 2009 08:50:06 +0000

crazy stuff. It kind of relates to the discussion we had in Boston though huh. Imagine if they cut off a Google server farm instead of Morgan Hill. We’d all be lost! heh.

Im definately pro re-population rather than archival of digital artefacts.

By: Daniel Tunkelang

Daniel Tunkelang — Thu, 23 Apr 2009 16:03:25 +0000

Wow, I'd gotten a 52K number from this article, but rereading I see that was just the estimated number of Verizon land lines. At 100K, that feels like disproportionate impact for the size of attack.

By: Roy Scribner

Roy Scribner — Thu, 23 Apr 2009 15:52:56 +0000

One thing we all (I live in Morgan Hill) learned very quickly is that an emergency stash of *cash* is essential. With the trunk line down, there were no operating ATM’s, or even debit card processing at retailers (grocery, gas stations, etc.). Also interesting is how all forms of communication were affected; landline, wireless, Cable, DSL, etc. I use an AirCard for redundency, but am now forced to rethink that selection.

btw – Morgan Hill (pop. 34,000) was just a sidelight in this outage, that affected around 100,000 people.

By: Milan Merhar

Milan Merhar — Thu, 23 Apr 2009 14:31:46 +0000

This event generated considerable discussion on Dave Farber’s Interesting People list (e.g. http://www.interesting-people.org/archives/interesting-people/200904/msg00100.html )
As said there, part of the vulnerability stems from the current Internet billing model and its resultant ISP bias towards internal vs distributed network resources — their own network infrastructure is expensed based on depreciation of the sunk cost, while peering with another carrier is an operating expense payable either in “cash money” or commitments of reciprocal transit (and carriers hate to commit resources to competitors.)

As for cloud computations, individual data centers have very high internal bandwidths and very low latencies, making it relatively easy to migrate computation from one resource to another. But, once that move spans physical locations (e.g. for failure recovery, or just to get a better resource/cost deal) your application’s data and stored state suddenly seems to have a very high inertia. It takes time to move Terabytes and someone’s billing you by the Gbyte for that I/O!