Ask HN: Estimating hosting costs

[pktm]

All,

Does anyone have experience or tips on how to estimate hosting, bandwidth, etc. costs for a web-based business model?

I'm working up a back-of-the-napkin business plan and would like to be able to throw something at the hosting costs column. The margins are somewhat tight at it is, and I know how these sorts of costs, specifically, can completely up end profitability.

I'm figuring about 15-30 MB of images per user, plus whatever it costs to receive those images one at a time (RESTful interface), then batch process them once a month (cron job + sending them off to some other RESTful interface).

Then, of course, there'd be some billing logic in there, maybe Stripe.

I'm thinking this would involve something like AWS, plus, I don't know, dreamhost, perhaps?

Sorry, I know this is vague (and that's part of the problem), but I've never estimated something like this before.

I'd really appreciate any help or suggestions.

Thanks!

[hhw]

Just determine how much the cost per user is, and make sure your business model works for that. Beyond that, I'd say not to worry too much about your total hosting costs as the reality is, it's impossible to predict how much resources you'll need until you're actually using them. Most of the time, people grossly overestimate how much hosting they really need on new projects.

Instead, I'd recommend focusing on being flexible enough so that you can start small, and upgrade your hosting as needed. Document your entire setup so that you can quickly re-create your environment when migrating to a new server. If your business model works when small with something like AWS, it's only going to work better as you scale up, as your costs will go down considerably as you get larger. You'll see a dramatic drop in costs once you outgrow cloud/VPS solutions and get big enough for your own, entire dedicated servers if you don't limit yourself to the really big players.

[dangrossman]

Unless you're going to run a million dollar TV commercial on launch day, chances are you won't have more than a few hundred to thousand users for a long while. From what you've described, a $20/month Linode VPS will be sufficient.

You can upgrade the instance's resources as needed with a few clicks of a button, and if image hosting is the primary use of storage and bandwidth, that can be offloaded to Amazon S3 for pennies while the VPS runs the website/API.

Plenty of successful startups have started just like that. Running the website/API on EC2 is where costs skyrocket. There is a huge premium there for the benefit of being able to programmatically start and stop instances. If margins are tight, you would do much better with a VPS or even a single dedicated box. I think Foursquare was well into the millions of users while still running their database on a single server with 64GB RAM.

[rush-tea]

I am also in the same stage as you. At first, i m thinking of shared web host such as godaddy or 1and1, and then watch the traffic daily. If it spikes to the point that you need a VPS, then you upgrade immediately.

Not sure if this way would be doable for you. But to me, why waste fund on something that you never use if you can allocate your fund to something else (i.e business cards, SEO, marketing, etc)? i remember reading that before pinterests got big, they only have few hundreds or low thousands users for the first few months they were starting.

Please let me know your thoughts on this.

Thanks.

[shail]

take a server (basic config or medium one) from linode, prgmr or digitalocean and that should do for initial year or so. Some image downloads will not land you in any trouble unless you are going to be pinterest in less than a year.

[threeseed]

Anything but Linode. People REALLY need to stop recommending them.

[orangethirty]

Whatever number you arrive to, double it.