Which do you use: Physical server, or hosting in the cloud? | Hacker News Reader

Which do you use: Physical server, or hosting in the cloud?

11 pointsfatalerrorx315 comments

I'm curious what other Startups are pursuing as a means of hosting their web apps.

I personally choose having a physical server at a location (I typically buy or build the server myself)

The basis of my choice relies on the breadth of configuration and customization that can be done when you have root access to a 100% dedicated server.

As a background: I'm currently working on a Startup that's in the Healthcare IT realm and the web app that we're building is very resource intensive and requires many different packages.

Now granted not everyone is in the same boat and doesn't require that much flexibility but I'm curious how many still do choose to have a physical dedicated server.

Eventually if the startup that I'm working on takes off, I will most likely need to use some sort of elastic hosting (Amazon), but this comes after you see whether your idea has legs and that you can raise funding required to build out the infrastructure required. In the meantime having a physical server cuts down on costs (no monthly recurring costs, just the 1-time fee for the server and $10/mo extra to be able to host off my Internet, and I have a fully 100% dedicated server). Best part about this option is if the Startup fails, the server can be used for the next endeavor and I'm still only paying just $10/mo for dedicated hosting.

Paying for a similar hosting setup that I have now would probably cost me several hundred dollars a month.

Please state your choice/preference below and any pros/cons that you've come across

We use both: Dedicated servers (at rackspace) running a private cloud for our website and shop plus various cloud based systems (EC2, Rackspacecloud, et al.) for testing and non-mission-critical stuff. You can create mission critical stuff on cloud servers, but you must "embrace failure", i.e. build the whole setup with failures in mind. We have an extensive post about our setup in our blog: http://www.paessler.com/blog/2011/11/15/networking-basics/fa...

Do you mind sharing pricing info for the private cloud? EDIT: NVM, I see 1,800.00 on the costs page per month...what does that include?

We're in between: full-on dedicated servers with SoftLayer for more intense tasks. We like being able to have other people work on the machine without having to drive to a data center in the middle of nowhere, but knowing that we have 100% of the hardware at our disposal (as opposed to virtualized disks and sporadic performance on EC2). It's really not that expensive either and we can still bring new machines online in a couple of hours if necessary.

I can still manage the server from anywhere in the world via SSH which I have setup -- of course if the Internet goes down at the server's location then all bets are off, but with EC2 do you have the ability to modify the apache configuration files? I need this ability for the web application that I'm developing

Yes, you have full root access to your EC2 instances.

Interesting, will have to look into this

I've always wondered though, is it as easy as copying the disk image on my current physical server and uploading it to Amazon? I would need the same configurations and packages in place and I'm just trying to understand if I would have to remember every configuration I ever made to get me to this point, or if it's as easy as a simple copy & paste

I don't think you can import an image into Amazon.

You will need to start with an EC2 image (AMI) of which there are many both from Amazon and 3rd parties. Some of those AMI's may have close to all you need. I personally started with a bitmami image (http://bitnami.org/stacks). Once you have all your configuration done on that you can create your own personal AMI from it and use it to create additional instances (dev, staging, different customers, etc). That AMI will work for creating different instance sizes (memory/CPU). But I don't think there is a way to export an AMI and use it outside of Amazon.

No I don't think you can directly import an image of your disk, but if you can find an Amazon image (AMI) of the same operating system you are using, you could probably just rsync your /usr/local/, web data, and database directories and get something approximate to what you are doing now. That said I think it would be worthwhile to start from scratch so you can at least have a fresh memory of how your site operates.

We currently use two hand built servers that are clones of each other behind a firewall. We've replaced them recently with much faster machines, sticking our databases and indexes on SSD's etc, and we take the retired servers and give them directly to the developers so they can build and test whatever they want in the exact same environment.

We buy physical servers in a managed datacenter. They provide the box and KVM access to it. We pay a reasonable price for top of the line hardware thats not on our balance sheet. We do all of the software bits (OS, updates, deploying software) and if the server has a bad drive or somehow else breaks the data center folks deal with it.

We use 2 physical servers at Rackspace sitting behind a Cisco firewall. They are mirror images of each other and they load balance between them. Should one fail the other covers us until both are operational again. Rackspace handles hardware issues, backups, and monitoring. We do everything else via remote access.

Sounds like this setup would be ideal for our purposes after we establish ourselves. I like having a non-shared physical server, however the downfall currently is that I don't have power backups or fail-safe backups should hardware begin to fail, that would all have to be done by myself which could become chaos if the product is successful, but I guess the hope is that if you receive that level of success you'll be able to fund raise to build out infrastructure and the team to maintain everything and build out further.

What are the specs on your servers and what kind of load can you handle? I'm curious because I just have a single server, the specs are ok, it's a quad core 2.4ghz processor, 2GB RAM (forgot the rating, but I think it's DDR2 and it's dual channel I believe) with a 1TB hard drive, runs Ubuntu Server (latest release), so it doesn't have a GUI sucking up any RAM, but I'm not sure what kind of load (# of users it can handle simultaneously) and would very much appreciate some insight if you have any stats available. I currently have apache configured for 600 Max Clients, but the web application itself uses a decent amount of memory because it utilizes website scraping...With 600 Max Clients and 450 started it shows that I have 700MB ram free, I have a feeling I'll need some more RAM before a wide launch, but I think what I have should be sufficient for a beta launch of several hundred users

And of course on a beta launch (it's a closed beta) not everyone will be on the app registering at the same time, so it would be probably at most 5 simultaneous users at any one given time

We're all on Heroku at the minute. Paying for one Postgres instance with API, Website, Admin and Push dynos.

Both, because you can't run OSX on a virtual server and we need them for CI with Xcode.

Reply on news.ycombinator.com