[AusNOG] census issues tonight

[Mark Newton]

On Aug 10, 2016, at 7:09 PM, Simon Sharwood <simon at jargonmaster.com> wrote:

> IBM isn't responding to anyone. the gummint won't say anything cogent. the talking points are clearly to say this is not an attack. If not, WTF is it?

Normal production utilization of a system that was capacity-planned by muppets who don’t know how to capacity-plan.

ABS said they tested 1 million submissions per hour for 8 hours, and briefly tested 150% of that.

They also launched a months-long PR campaign telling everyone they’d be fined if they didn’t submit, so I reckon they had 10 million households trying to submit between 7pm and 9pm after they finished dinner.

That 2 hours turns into 5 hours when you smear it across several timezones. 

10 million households in 5 hours is 2 million submissions per hour if it’s constant. Capacity plan for at least 4 million per hour to deal with burstiness. Expect a big peak at the beginning of the time window, because Brisbane, Sydney, and Melbourne are the most populous cities and are all in the same timezone.

Even if my estimates are out by a factor of 2, the ABS’s smug self-satisfied preening about their half-million-dollar load test were woefully underprovisioned.

Pick apart my numbers above: See if you can find an error I’ve made or a point of disagreement that’d enable the ABS’s 1 million per hour figure to be good enough.

This wasn’t a DoS, this was ten million Australian households doing what they were told to do, on a system that was designed by people who haven’t the faintest idea what they were doing.

  - mark


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20160811/1672a148/attachment.html>