3 ms·
Missed the real answer: In your cage at the local colo. The sweet spot for the Microsoft stack is one beefy server. While you can rent that from a cloud provi
by jasonkester 13y ago
Missed the real answer: In your cage at the local colo.
The sweet spot for the Microsoft stack is one beefy server. While you can rent that from a cloud provider, you're just plain always going to get a better deal by building your own box and dropping it in a datacenter.
It's not anywhere near as daunting as it seems to do so, and it will run you about $400/month all in, amortized out over a 3 year box lifetime even including licenses.
That'll get you the firepower of a medium sized cluster of Linux gear running your open source stack of choice, without having to deal with a dozen virtual machines spinning up and down to deal with load. It'll just handle as much traffic as you can throw at it, up to the point where you're big enough to simply pull a half dozen of your hundred devs off what they're doing to help scale you up.
That's probably the reason you don't see many .net paas companies around. For the most part, they're not the right tool for the job.
- skeletonjelly 13y agoSome people just don't want to deal with colo. I know I'm sick of it. I enjoy mucking around with the stuff and spending the company money of course but when shit hits the fan it's pretty annoying. Same could be said for cloud of course. Currently dealing with a random NIC lockup on a VCloud hosted SQL Server for a client. I feel hosting issues are inescapable regardless of where you go.
- rbanffy 13y agoThat's a good solution, until you need to do a major upgrade, or when you need to hire additional capacity for a temporary spike (say a campaign). It's only a solution if you are OK with occasional downtime - because you'll need to stop everything to apply service packs. With something like EC2, you just spin up a similar box with newer patches applied, redeploy your application, load-balance between your versions and just delete the old one when you are happy. I cannot imagine going back to colo. Or even physical boxes. If I were forced to have a physical box, the first thing I'd do would be to deploy everything on virtual machines or containers inside it.
- jevinskie 13y agoIf you're willing to spend the cash, 3 servers can offer you a hot failover with a spare staging server. Of course with just 3 servers, you would have to accept failing over to a different version while you update the 2nd server.
- esw 13y agoYeah, it's a cost vs. peace of mind equation. I've done both colo and managed, and in my case, having someone (e.g Rackspace) manage the ecosystem is worth the added cost.