"10 Real World Things to Consider Before Scaling Your Startup"

Damian

http://readwrite.com/2012/12/31/10-real-world-things-to-consider-before-scaling-your-startup

Thought it was a good article, could probably help a few people around here.

miTgiB

"10. Estimate Growth, Then Divide By 2"

lol, I do that, then take 10% of the result

HalfEatenPie

Estimate Growth eh?

I would aim for less than 5% of projected growth.

dano

I have estimated growth a few times, and was only wrong once -- essentially, the marketing department that was handling the site I was asked to scale, didn't include me in their plans to introduce the site to the public two weeks earlier than I was told, and also bought a campaign on a site that is in the top 200 site on the net(=huge traffic) .

Needless to say, I whipped a u-turn on the 101 and headed back to scale the site with a couple of nodes behind an nginx load balancer - what was once a single Apache 512Mb instance and a 256Mb nginx instance, turned into (4) 2B nodes that needed to run the site during this campaign.

I would say that you should take a few considerations into account when scaling a site:

1. Current load on the system
2. Anticipated load on the system (product/marketing thinks they are going to sell 1Billion in the first day, dont let them convince you to over commit resources, but be ready to).
3. Software stack - best practice methods for dealing with different environments (wordpress blog going viral, cart campaign, etc). You should research how others are scaling their environments with your current software stack, and take some notes. If your trying to scale someones homegrown app, please communicate with them regarding their experience with the software. I have found that most developers thinking of scaling something they built themselves, usually over allocate resources, so you may* have to bring them back to earth with their ideas of 25 nodes running their BRAND new site, that has NO current marketing or traction. Also, you may know better than they do, in regards to expire times for web objects, caching, etc...use this knowledge to help them scale cheaper and more efficiently.

At the end of the day, I failed because of "lack of communication" with my client regarding their plans - although we only lost an hour, the campaign they ordered cost them 10K US dollars per DAY to run - so one hour of downtime(with lost sales, campaign cost, human cost, virtual machine cost, my labor cost)....was totally avoidable.

Ymmv - thanks for reading.