Virtual Private Server Hosting vs. Colo + Own Server(s) for Early-Stage Web-App

Question

I'm working on a web-application and need to figure out the best way to host it. I'm looking to minimize "Total Cost of Ownership," but it's more important to keep hosting costs low than labor costs in my case.

Potentially, some day, it will grow quite large (in terms of user-base) with heavy traffic but, of course, it will start out pretty small.

So, does it make sense to start out with a low-end virtual private server hosting plan, grow the server slice until it stops being cheaper than running my own servers, and then buy servers and set them up at a colocation provider, with all the pain of migrating everything,

-OR-

Buy my own server, set it up at a colo provider, and waste most of the server capacity to start, but reduce the pain in scaling from small to large significantly?

Has anybody tried both of these options? Are there pros and cons beyond the ones I'm considering?

Thanks.

Addendum: Thanks for the suggestion, but a shared hosting account probably wouldn't cut it, since I need to run persistent processes and access things like sendmail configuration files. I'm looking at either a VPS or my own server(s).

Answer 1

I would highly recommend going with a VPS account. I've had clients who have spent many thousands of dollars and countless hours trying to get a perfect dedicated hosting environment setup and then never came close to needing it's capacity.

With VPS, as long as you use the virtualization technology that you want to stick with, it should be very easy to migrate/replicate the whole "server" to a new dedicated box when needed.

If you need SQL Server, I'd recommend starting with a Shared SQL Server rather than purchasing a full SQL Server license or using SQL Express.

Answer 2

In the beginning you will actually find that scaling a VPS solution is far, far easier and more flexible than scaling on physical boxes. Personally I use and would highly recommend Slicehost. You can create and destroy new slices on the fly to test out new ideas or prototype something and you're only billed for what you use. You can also instantly scale a slice by upgrading to the next level up; within minutes your box is up with the new CPU and memory allocations and the cost is prorated.

I recommend Slicehost because they are developer-centric and are really a great bunch of guys.

Once you grow large enough to outgrow your VPS farm (which will take a while, trust me, these VPS's can be pretty beefy these days) you'll have enough cash to worry about moving to a physical platform.

Answer 3

Start small. Yes, some scaling headaches, but if you want to avoid them completely, use a Platform as a Service (cloud) service provider instead of hosting it all yourself.

Statistically, most projects fail, and the added overhead costs are not going to improve your chances. Even very popular and large services have difficulty making money. Hosting companies largely make their money out of the gap between the dreams of a customer and the reality that follows.

So, pragmatically, start hosted/shared so you can concentrate on the technology you're developing, rather than on maintaining a remote box, and ascend the tiers as the project gains users/customers.

I don't speak from experience, though, except for having dreams that don't get very far.