Hacker News Reader

Comment by iampims | Hacker News Reader

We `reset` the blog a few months ago. That’s why there are only a couple of posts.

Sorry to hear the news. I hope you and the team go on to new and greater things.

While it’s heartbreaking to have to shut down something you’ve poured your heart into for many years, we have learned many things along the way that will help us for what’s coming next. Thanks for the kind words.

Any reason why the export period is so short? The archive team is trying to make a backup of the site at the moment.

Any help would be appreciated on EFnet IRC channel #firespring or #archiveteam

Archive team projects: http://archivingyoursh.it/Current_Projects

@pinwale Starting March 31st, no new content will be created on Formspring. The site will be in read mode only. Ping me tim {at} formspring.me if you want more info.

Can I ask how much (low/mid/high x-digits) the raw AWS hosting costs of a site like this are?

I'd love the idea of trying to take over something like this, but I have a suspicion I might be in way over my head...

I can’t disclose the costs but I can share a few things:

- Our cassandra and Redis clusters are/were the most expensive pieces of the architecture.

- the PHP front-end wasn’t really efficient and required many `medium sized` instances.

- Sites with huge fanout effects are expensive to run.

- Multi-platform : iOS, Android, "touch" and www. Maintenance gets expensive.

- Multi-AZ on RDS comes at a price

- If you intend to run on AWS, Reserved Instances are your best friends.

I still don't understand why anyone would want to use AWS unless you had money to burn. I haven't followed the prices as of a couple years ago but I'd say most people on here are smart enough to find cheaper solutions that can achieve the same thing. If saving money is your goal you can make it happen--there are providers out there with the right APIs that allow you to build something that can scale (or adapts) at a fixed/linear cost. If you have the money and it isn't worth your time to setup the infrastructure then AWS would be the right decision. I just don't understand the infatuation.

The real benefits of AWS is scale on demand, if you need it.

Formspring reached 1 million users in 45 days. How do you plan for this on typical hardware? Especially when you had no idea ANYONE would use your little side project?

AWS isn't cheap. It offers building blocks upon which you can "easily" build and scale a website. Not for everyone, but good enough for us.

Do you always have to operate expecting sudden spikes and that prevents you from moving to a cheaper hardware? If traffic gets over what you could serve, display 'our servers are busy' message and perhaps that way you could have kept the service alive if costs halved..

> Do you always have to operate expecting sudden spikes and that prevents you from moving to a cheaper hardware?

In short, yes. This is inherent to the nature of Formspring – or to some extent social networks.

Since most of your actions on the site are being `broadcasted` to all your followers, and since the distribution of followers follows a pareto distribution, it is very difficult to predict load at any given time, as one single action by a very popular user could trigger hundreds of millions of events (publishing to everybody's stream, sending emails, push notifications, updating counters, spam filtering, etc.)

We benefit greatly from the elasticity of AWS, and `scale up` accordingly to deliver this event in a timely manner.

If we were not concerned by `delivery time`, we would be able to smooth it out, but at the expense of engagement… and as you can imagine, engagement in a social network is quite important :)

I think the plan is that if you have to manage costs as closely you describe (as you or I might), your revenue opportunities are too small to care about in comparison with the founders' expectations of a giant exit.

Many of us are not even playing the same game.

Reply on news.ycombinator.com