Fri 30th Mar 2012 @ 21:46 2012: Efficiency

The BBC have posted an article entitled Go figure: Does efficiency always save money?. The article makes few direct claims, but it does say that we are spending 63% more on IT now than in 1997. We are getting much more than 63% more use out of our IT than we did 15 years ago, too, so the fact that we spend more is not really the relevant factor.

I have taken figures from the graph on the BBC website; I might not have it quite right, but they suggest that we spent around £3,800 million in 1997, and £6,200 million in 2010. (I've gone out to 2011 for convenience, as Moore's law works in 24-month chunks). That is a higher spend, but we're spending more on it because we're getting more out of it.

If you had a £1 lottery ticket guaranteed to pay out £1.05 for every £1 you put in, from an initial budget of £1.00, after 20 weeks you'd be able to spend £2 and buy two tickets. Of course you would - you'd be making a 10% profit each week instead of a 5% profit. 10 weeks later, you'd be able to buy 3 tickets a week, for a 15% profit.

Moore's Law is even better than that. The BBC article claims that Moore's Law claims "Every two years you can buy twice the processing power for half the price." Actually, it's that you can buy twice the power for the same price. But it's still pretty good.



If we liked our £2,000 PC in 1997, with Windows 3.1 and MS Office 4.3, we really like our laptop, netbook, phone and maybe even a tablet now; the least powerful of these cost a fraction of that 1997 PC, and is far more powerful, and makes us far more efficient.

That is why we are spending more on technology - the more of it we have (up to a certain limit, I'm sure), the more effective we are.

From the figures above, even if our spending stayed at 1997 levels, we'd have IT power 12 times more than we had then. By spending more on increasingly efficient technology, we've actually got 77 times more power than we did 15 years ago.

So - just because we're spending more on something does not mean that it is less efficient than it appears. Rather the opposite - the more useful something is, the more likely we are to keep on using more of it.

Wed 21st Mar 2012 @ 00:36 2012: Traffic Flow

The Technical University of Dresden has a Java applet modelling various types of traffic flow at http://vwisb7.vkw.tu-dresden.de/~treiber/MicroApplet/. Fascinating to someone who spends 20+ hours a week on motorways; boring to someone who drives 2 hours a week on minor roads. As someone who spends 8-10 hours a week on motorways, I find this an interesting and educational toy.

It's interesting to play with the different scenarios; for example, http://vwisb7.vkw.tu-dresden.de/~treiber/MicroApplet/OnRamp.html tells us that "At 1600 veh./h, no breakdown occurs, but an existing jam will not dissolve" - so once there is a jam, you need to reduce the number of vehicles entering the affected system. I'm not aware of any measures (other than Radio and SatNav telling drivers to avoid the area) trying to control this.

Wed 7th Mar 2012 @ 15:40 2012: Unwanted ssh-agents

If you include an eval `ssh-agent` in an X session startup, in your ~/.profile, or just run it interactively when logged in to a server, you could end up with quite a few unused ssh-agent processes over time, because your shell disowns them and they are reparented by the init process. This does no real harm, but it can make the ps output messier than it needs to be.

Adding this to ~/.bash_logout would kill off your no-longer-needed ssh-agent instance:

SSHPID=`pgrep -x -u $UID ssh-agent | grep $SSH_AGENT_PID`

[ ! -z "$SSHPID" ] && pkill -9 $SSH_PID

echo $SSH_AGENT_PID | grep "[^0-9]" > /dev/null || kill -9 $SSH_AGENT_PID