Someone did the math

April 24th, 2006

[…] there’s the question of the rate at which we’re all going to be put on this thing – I’ve done some calculations before on the speed at which the system will have to work based on 700,000+ people turning sixteen every year for ever, which showed that for normal office hours there isn’t actually very long to do all the ‘background checks’ and duplicate biometric checks that would be required for a gold standard database. Now I’ve seen the first official figures for the rate as estimated by the S.E. and his hapless sidekick Andy ‘Noddy’ Burnham, and they’re alarming from the point of view of a professional IT worker or indeed anyone with a basic mathematical knowledge:

About 80% of the UK population has a passport and all will have to be renewed within the next 10 years, at an initial rate of about 7 million people a year, a Home Office spokesman said.

Now, even leaving out anyone *wanting* to go on the damn thing or people over 16 getting their first passport (who’d have to go on it, like it or not) or the several hundred thousand foreigners coming into the country whom we now find will be put on it* (does this mean another fully manned registration centre at every port of entry, working round the clock, or do you just trust them to roll up of their own accord after entry?), we’re looking at a system workload in the first year of ten times the long term load. In a new system set up by a Government IT contractor, paying the kind of wages for operating the kind of systems that have led to such high morale and efficiency in organisations like the CSA that’s a hell of a task. This is all supposed to kick off and be working at that rate in 2008. Assuming December 2008 (and December is a really bad month to launch new IT kit for myriad reasons) that’s 32 months away.

Let’s examine what 7 million NIR entries per year looks like:

Days in a year – 365 (ok, I know 2008 is a leap year!)
Weekends – 104 days
Weekdays – 261 days

Public Holidays: usually 8 days a year or so including time off around Christmas

Working days for NIR per year – 261-8 = 253

Registration centres : 70

Number of registrations per year per centre : 7m/70 = 100,000

Per centre per day : 100,000/253 = approximately 400

Working hours of centre – well, since you’re forcing people to come along you can’t make it the middle of the night, so say 9 to 5 inclusive – 8 hours per day (or 8*253 = 2024 hours per year)

So adding it all up, from NIR Day 1 for ten years you’ve got to keep processing people at the rate of 50 per hour at every centre, or one every 72 seconds, each of whom requires a scan of the whole central NIR to avoid multiple registrations, so the database has to be up and accessible every minute of the day to avoid delay.

In the early days it’s a nailed on certainty that we’ll get failures, resulting in potentially hundreds of people making pointless journeys (say it’s down for an hour during a particular day – that’s 50 people at each centre having their time wasted, a total of 3500 people). I have no idea of the MTBF for major government IT projects, and they almost certainly won’t tell me on the usual ‘commercial confidentiality’ grounds. What I can do is provide some figures based on possible percentage reliability and estimate the number of people inconvenienced per year and the kind of reliability that would be required *from day one* to stop the scheme sliding into chaos.

Reliability (uptime during working hours) People inconvenienced Time offline in a year
99.999% 71 73 seconds
99.99% 708 12 minutes
99.9% 7084 2 hours
99% 70,840 20 hours
95% 354,000 101 hours
90% 708,400 202 hours

I’d suggest that anything much below 99.9% reliability is going to be seriously political in terms of people claiming loss of earnings, loss of holidays etc. 99.999% is cloud cuckoo land for a scheme of this complexity built in 32 months. Not a lot of margin for error between those two really. You reach the million people inconvenienced per year mark at about 85.8% uptime, by the way. […] 

Leave a Reply

You must be logged in to post a comment.