qv the Bill Hicks routine where the waitress asked him "Whut yew readin' for?". Took him a moment to realise she hadn't asked what was he reading, but what was he reading for...

According to the back of this envelope:
660000 oil barrels spilt = 104931615 litres
At 17500 l/m = 5996 minutes.
At 24 * 60 min/day = 4.16 days.

So just one of these collectors could have hoovered up the entire spill in well under a week in perfect conditions. Even 10% of efficiency is still only six weeks. Even 10% efficiency and only working in daylight is still only three months.

We estimate savings of ~30% of server cost, none on IT staff. If you do it right. If you do it wrong, costs can rise.

I see your point. It's not my impression that moving things 'into the cloud' will magically make any complexity go away. But it does make that complexity easier and easier to manage.

We can stripe servers around the world at service setup time with one command.

Load balancers now intelligently remove defunct machines and increase capacity (from other zones) to compensate, provided you have defined how you want them to act (and no system will ever define that for you for a known cost).

It's not a panacea for all IT ills, but it is strong medicine.

But they weren't down. I have servers in Dublin. I also have striped redundancy across other EC2 datacentres. We had 100% uptime last night and I'm now watching the Dublin based machines recover gracefully.

Yes, I agree it's "advanced virtual hosting with a different name". But it didn't break its promises.

Oh silly underinformed person. There is a datacentre in Dublin. It is one of six Amazon datacentres. The others were unaffected, as were (our) public facing services, because only some (of our) servers are placed in Dublin.

It looks like a cloud from the outside. Those of on the inside know where the servers are because we want to choose where we place them for latency / redundancy reasons.

Rubbish. We stripe across two EC2 zones. We would always want at least two servers anyway in case of any server failure. We still had 100% uptime over last night because failover worked correctly.

We are now seeing servers come back up gracefully and so far have not had to take any remedial action - a watching brief.

And it's cheap.

Back when I was a wee lad, my wonderful maths teacher taught us to consider aleph-0, aleph-1 etc values of infinity as follows...

There is an infinite number (A) of lines of constant y that don't intersect y=0.
There is a larger infinity of lines of non-constant y that do: For each value of y at the y-axis (an infinite number equal to A), there is an infinite number (B) of angles at which the line can be drawn. Thus the first infinity cancels out, giving the probability of an arbitrary line not intersecting y=0 to be A/AB == 1/B == 0 because B is infinite.

Alternatively: for any location in 2d space, there is one line that does not intersect, and an infinite number that do. Probability of picking one that doesn't intersect is therefore zero even though there is a line that doesn't intersect!

qv Hilbert's Paradox.

Justin.