Provisioning for a Million Eyeballs

There’s real value in inviting industry people to talk about the issues they face in the real world, and it’s obvious Prof Ben understands the value in this.

I was lost for the most part of the lecture, but i think i managed to take away a few concepts in the end.

I liked how Zit Seng used the examples of massive events like the IT show, Singapore Grand Prix and his particular example regarding the hotel lunch scenario. How the heck does one plan for 10,000 people coming in simultaneously, and, for example, going for lunch? Many issues arise: how to ‘guide’ them in an orderly manner to the desired place, such as which direction the queue for food is supposed to be? Also, for the IT show, how do the organizers provision for thousands and thousands of people thronging the place at the same time, all coming in different openings?

I think one key thing to take away from the above examples is that no matter how extensively you plan on paper for how things should flow in real life, things are probably not gonna turn out exactly as planned on paper. This was nicely summed up in his point: “Many webapps work perfectly in development and QAT, but fall apart in production”

For example, when Universal Studios opened their online facilities for online booking of tickets, their system still failed miserably amidst huge public discontent. Why does such a huge organization like that, with their millions and millions of dollars in resources, still not manage to keep a seemingly innocent online ticketing system, up and running? This is perhaps put into words very aptly by Zit Seng in his slides: “Need to simulate realistic user activity”.

Another main point i brought away that lecture was that how often the following scenario occurs: when an application breaks, the software people are gonna blame the hardware people, and vice versa; the applications people are gonna blame the network people, and vice versa. Why does this happen? The “Myth of the Overload” slide addresses this. Probably, many a times people are just gonna upgrade and add better servers to fight the fire. However, they are not addressing the root of the problem, which Zit Seng explained the technical side of very well. Shall not try to regurgitate what he taught, as i do not have the know-how to do so

Key takeaway here: get to and understand the root of the problem before going about solving it. This way, you’ll reduce the chances of the same problem occurring again under future heavier load. Same goes for similar situations everywhere else. Common sense stuff, but we tend to not apply common sense when we ourselves are put in these situations.

Advertisement