Starting a Shared Web Hosting Company
-
We were a web host going back to the 1990s. Back when there weren't the low cost hosts out there everywhere. It kind of made sense then. We phased it out over a decade ago, it was just not a viable business model until you have tens of thousands of clients and even then, very hard to do.
-
One of the difficult things is that to have good pricing you need to run on your own hardware. But to have scalability you'd want cloud. That makes things very hard.
Likely your only reasonable option would be to start with dedicated hardware and project as best as possible when to buy and just stop taking on new customers if you grow too fast until new gear arrives.
Something like Dell R730xd with some SSDs for the MariaDB server VMs and then SATA disks for your application VMs would work well and be cost effective.
Something like the Scale HC2100 and HC2150 hybrid tiered cluster would do the same and allow you to grow from three nodes to a dozen or more with high availability built in and growth built in giving you an HA infrastructure plus a growth system. Otherwise growing past one node is going to be very, very painful as you move from local storage to a CEPH cluster or whatever.
-
A HyperConvereged system (compute and storage in one). So you can just add another node a scale out both storage and compute power you just add another node.
However this is not cheap, a single node with a few TB can easily cost as much as house.
-
@Jason said:
A HyperConvereged system (compute and storage in one). So you can just add another node a scale out both storage and compute power you just add another node.
However this is not cheap, a single node with a few TB can easily cost as much as house.
A very cheap house
I know that an entry level Scale system starts at roughly ~$25K in the US. That would be a full three node HA cluster with all of the storage and compute included. So while it is far more than just buying a single server and figuring out scaling later, it gets you into the HA world and the scale out world all in a single purchase and let's you host quite a lot of web before needing to invest more. And it would be under $300/mo to host in a Tier IV datacenter.Scaling to more capacity would be done in chunks of like $8500 or so. So you can grow as you bring in customers.
-
How does bandwidth work in those solutions? I hear Scott saying that he gets 100 Mb connections in those colo's but I'm assuming that it's just one per server. Not sure if a Scale cluster would be considered 1 or 3 (assuming 3 servers?) But really that's not relevant - what's relevant is what you are providing to your customers on that cluster.
I have no clue, so I'm putting this out there as a question.
-
@Dashrender said:
How does bandwidth work in those solutions? I hear Scott saying that he gets 100 Mb connections in those colo's but I'm assuming that it's just one per server. Not sure if a Scale cluster would be considered 1 or 3 (assuming 3 servers?) But really that's not relevant - what's relevant is what you are providing to your customers on that cluster.
Depends if you are buying by the rack (this would be a quarter rack probably) or by the "U". If you buy by the U, you get one connection from each node. If you buy by the rack, you get one connection to your rack.
-
You could easily negotiate a quarter rack with three node, two switches (that's 5U) and a GigE drop with metered bandwidth.
-
@scottalanmiller said:
@Jason said:
A HyperConvereged system (compute and storage in one). So you can just add another node a scale out both storage and compute power you just add another node.
However this is not cheap, a single node with a few TB can easily cost as much as house.
A very cheap house
I know that an entry level Scale system starts at roughly ~$25K in the US. That would be a full three node HA cluster with all of the storage and compute included. So while it is far more than just buying a single server and figuring out scaling later, it gets you into the HA world and the scale out world all in a single purchase and let's you host quite a lot of web before needing to invest more. And it would be under $300/mo to host in a Tier IV datacenter.Scaling to more capacity would be done in chunks of like $8500 or so. So you can grow as you bring in customers.
Most be pretty low end, VBlock starts around $180,000 for entry level. though for true shared webhost only you don't need much the fanciest.
-
@Jason said:
@scottalanmiller said:
@Jason said:
A HyperConvereged system (compute and storage in one). So you can just add another node a scale out both storage and compute power you just add another node.
However this is not cheap, a single node with a few TB can easily cost as much as house.
A very cheap house
I know that an entry level Scale system starts at roughly ~$25K in the US. That would be a full three node HA cluster with all of the storage and compute included. So while it is far more than just buying a single server and figuring out scaling later, it gets you into the HA world and the scale out world all in a single purchase and let's you host quite a lot of web before needing to invest more. And it would be under $300/mo to host in a Tier IV datacenter.Scaling to more capacity would be done in chunks of like $8500 or so. So you can grow as you bring in customers.
Most be pretty low end, VBlock starts around $180,000 for entry level. though for true shared webhost only you don't need much the fanciest.
VBlock is crazy stuff, doesn't have any low end offerings. Most hyperconverged starts at a fraction of that price. VBlock is, AFAIK, the most expensive offering on the market, not "normal" by any stretch. And only comes in very large sizes.
Other hyperconverged players like Scale, Starwind, Nutanix, Simplivity start around the same range as each other.
VBlock makes their money off of people buying the name, you pay a massive VMware tax on that gear. Not that that is bad, they have great technology, but like Cisco, you pay a ton of overhead just for the name.
-
Why delete this? If you have the ability to do it, more power to you. It's also a good reference for people who might be wanting to do the same thing.
-
@johnhooks I didn't delete it.....
-
@aaronstuder said:
@johnhooks I didn't delete it.....
Somebody did, and then you put it back up. When most of us were reading this, OP was deleted.
-
@aaronstuder said:
@johnhooks I didn't delete it.....
What it back?
-
No idea why the OP was deleted. But it is not purged, or was not last that I checked. That means that it can be recovered.
-
OH I see, it is ACTUALLY back already. Weird.
-
@scottalanmiller said:
OH I see, it is ACTUALLY back already. Weird.
That happened about the same time I posted about electric vehicles in the "What are you doing now" thread. Thought it had disappeared for a minute.
-
@aaronstuder said:
Going to start small and grow slowly focusing on providing excellent customer service.
For Operating System I'll be using CentOS 6/7.
For Control Panel I'll be using CentOS Web Panel (I think.....)
For Redundancy - I'll be using 2 providers and 2 different geographic locations (maybe as many as 4)
The website will be running on WordPress, and our ticketing will JIRA Service Desk.
Basically, I will be setting up this twice
https://assets.digitalocean.com/articles/architecture/production/production.png
Any recommendations?
What is your personal goal out of curiosity? Is this for practice, something to do, or are you actually attempting to make money?
-
Web hosting would be really difficult to make a decent living in to justify the work. I would personally focus on delivering other IT services like setting up networks/troubleshooting/etc. Remote and on-site to start. Then you could move into all remote down the road if you wanted and the business grew. Possibly hosting your own Colo if businesses wanted their gear offsite and cloud costs weren't feasible or possible with customer contracts. That's where I see the real money at, and overall the best customer service experiences. When you deliver a solution and people are in a panic because their network is down, you get big money and big thank you's.
-
Without reading the thread, as someone who started a shared hosting company in 1998 and has been dealing with it since then, it barely pays for itself. You really need to have tons and tons of customers for it to work out. Customer service doesn't really mean a damn thing, because the kind of people who really obsess about that when it comes to web hosting don't know anything, which means because they don't know anything, they'll go with GoDaddy or whatever their registrar provides.
The only way to really work is to provide hosting of something most others don't, like Windows hosting. If you're providing LAMP, your service has been done, and is everywhere, and nobody is going to move to you. Providing node.js hosting and other things is going to be much more difficult in a shared environment as well, so I wouldn't count on grovesocial moving to you. You maybe can get some of your clients to move their sites, but you won't be living off it.
Customer service couldn't hold up any of the many open source companies as making them "different", there's no reason to think it'd make a difference with yours.
I used to reply to all the threads on Spiceworks about this, every few weeks somebody else wants to do it. Weirdly no one ever comes back saying they were successful.
Also with your image/design. That won't work in shared hosting environments, because you don't know if the customer's application even can handle load balancing, most can't and it'll just confuse them and possibly break their app, or at the very least end up logging your balancer's IP address as every single one of their visitors.
The way we do it is having several Apache instances running on each server along with some customised stuff going on, chroot, etc. We do provide services where people can have load balancing, but nobody provides direct, out of the box load balancing or redundancy to customers who are looking for shared hosting, because people who are looking for that don't usually use shared hosting, and the kind of people who use shared hosting are the kind of people who don't know how to deal with it.
Security is a damn nightmare. I've seen many shared hosts over the years get rooted or have processed spawned from PHP, Perl, etc which worked outside of the configuration bounds, etc.
It'll cost a lot and you won't make your money back, unless you figure out how to be very niche, and then you've got other problems because if it's that niche there won't be a "how to setup X hosting company" tutorial out there, complete with dealing with billing, refunds, security nightmares, etc.
-
@aaronstuder I know I may be quite late to the game but the other option companies are doing is using the hyperconverged solution from HTBase (www.htbase.com)
With that, you can build a cloud environment utilizing existing or new hardware that you may acquire as well as it comes with an "AWS like" panel where your clients can go online, create their own virtual machines and have total control of that.
Along with that "AWS like" panel, called Fortis, you get an Application Market, where you can provide customers with customized applications that can be installed automatically as they create their vms. There is a charge back mechanism and others.
And, if you need to scale and want to use cloud servers, there is OCH, the OneCloud Hypervisor, that allows you to get resources from cloud providers such as AWS and GCE and make them work as internal infrastructure for you (www.htbase.com/och)
