December 11, 2008

550 words 3 mins read

“Cloud” versus Hardware – Costs

It really humors me how much news I’ve seen as of late about “Cloud Computing”. Granted I’ve gotten more into that specific niche, so, I’m reading more into it. The one thing that a number of “major” news stories tout is how much money will be saved by using the “Cloud”. News flash… Only if you use the cloud properly. To use the cloud properly, you need to scale up and down. That is really about it. It isn’t difficult, but it isn’t necessary for a lot of people.

Take my personal hosting for example. This blog and a number of other sites I own (or friends own) are hosted on a VPS, which I pay a very minimal amount for. I’ve gotten Dugg before and that totally hosed my server for about 12 hours or so. It would have been nice to have a cloud all setup ready to scale at a moment’s notice, as I could have held up perfectly through the storm. But it would have been stupid. A) It was one day out of several years. B) I’d be paying a lot more for even a small instance at Amazon.

So, say you have a reasonable website that gets a decent amount of traffic. Does the traffic spike? When it does, will you lose money for being down or slow? Do you actually care that much about the hour a month you traffic is increased? Cloud versus Hardware is just like leasing versus buying a car. The way I figure it, you can buy a server from Dell which is approximately equivalent to an AWS instance, and pay it off in about a year (at the AWS Instance cost). So if you are only running one instance 24/7/365, at the end of a year you’ve spent a bit of money and got nothing. If you own the single sever that is running 24/7/365 then at the end of the year you’ve spent a bit of money, but own a server. Now after that first year, the hardware is essentially free. Can’t complain about free, now can you? Of course there are upkeep costs like hardware failure… but that is another argument for another day.

Regular usage spikes are what it is all about. Keyword being REGULAR. You can look at CNN on 9/11 for a great example. Yes, their website got hit EXTREMELY hard and if they had been using the Cloud, they could have weathered it no problem. But is 3 days out of several years worth the costs? No. Not by a long shot.

Take another example, Woot.com. They get about 50% of all the purchases made for a day, in about 6 hours, as seen here. Every day at midnight, there is a rush of people which tapers off extremely quickly. Then, in the morning it picks back up again and then slides off and holds at a low but constant level. At midnight and 8am they can pay for 10 times the standard number of servers they need, for an hour. After that they can reduce the number of servers.

In the end, every instance left running in the cloud 24/7 is money wasted. Of course you need some base to keep your site going, but just keep the money in mind.