Page 1 of 2

Starting a hosting company

Posted: Wed Mar 07, 2007 7:57 pm
by alex.barylski
I'm not interested in hearing how any of you have tried and failed or gave up because the trouble wasn't worth it. So please no comments like that.

What I would like to hear is how I might go about setting up an actual hosting company. I mean:

1) Builging Security
2) Environment control
3) 24/7 emergency service

And so on...

I have a building already in sight, which offers all the above but how would I get connected to a major Internet backbone? My local service providers are the local phone company (DSL) and Shaw cable (national carrier - Cable).

There are of course others, but I assume they all share the same backbone.

I've called Shaw and asked but they couldn't/wouldn't give me an answer.

How do I go about finding out where to start?

As a developer, what kind of features/service do you look for in a host? Full control over PHP and MySQL is a shared environment good enough?

Cheers :)

Posted: Wed Mar 07, 2007 8:56 pm
by Ambush Commander
Well, the most expensive part of being a hosting company is, as you've begun to realize, maintaining the data center. I suppose the Wikipedia article offers some good pointers on what a good data center should have.

Regarding connectivity, if none of the major backbones service your locale, you're really out of luck. For a hosting company, having DSL as your line out won't cut it. Heck, their ToS probably doesn't permit you to host content.

Posted: Wed Mar 07, 2007 11:02 pm
by alvinphp
How about start with an established Data Center that provides you with the space, power, security, redundancy, and backbone. You can then concentrate on building the servers. I suggest you check out webhostingtalk.com, this is probably the biggest site out there when it comes to any type of hosting.

Posted: Wed Mar 07, 2007 11:03 pm
by Luke
why not start out as a reseller and then when funds etc. permit, you could make the move to a data center.

Posted: Thu Mar 08, 2007 3:31 am
by nickvd
What you described is running a data centre, not a hosting company. The best thing to do is to colocate some servers in a data centre somewhere (local if possible). Then you can move up to a half-rack then a full, perhaps a cage. That's when you should start thinking about starting a your own data centre :D

Posted: Thu Mar 08, 2007 4:05 am
by Jenk
I'd wait until I have a few or even several cages under my belt before starting my own datacentre :p

Colocation is the way.. if it's anything like here in the UK, there are certain areas available for hosting, dictated by where the main internet links are. (Redbus in the London Docklands area is the main UK hosting site.)

It's also a *lot* of work. 24/7/365 call/pager support, etc.

Posted: Thu Mar 08, 2007 5:18 am
by Chris Corbyn
What should a good datacenter have? Hmmm...

* Multiple ISPs with fat pipes to choose from (i.e., not biased towards one ISP)
* Huge UPS systems
* Generators that kick-in when the UPS is triggered (UPS doesn't run for long)
* Plenty of space to get around both the front and back of your rack
* Constant monitoring of life over the network
* Damn good air-con. If you walk into a data center and it feels warm, don't put your kit in it!
* Compressed gas fire prevention
* 24/7 security with fob-access

It becomes incomprehensively expensive. You have to consider that not only may just get power cuts (the electrical installations themselves cost a fortune!), you may also get fires etc and they need to be stamped out asap so a few fire extinguishers won't cut it.

Posted: Thu Mar 08, 2007 6:52 am
by onion2k
d11wtq wrote:It becomes incomprehensively expensive.
This. If you don't have an investment of £millions to hand then you won't be able to compete in the datacentre market. That said, a hosting company doesn't need it's own datacentre. Rent space in one.

What you will need instead is one of two things. Either awesome Unix admin skills so you can automate absolutely everything and run it all very cheaply, and thus compete in the low price market, or else you'll need fantastic sales skills to persuade people that you're worth spending a bit more money on.

Frankly it's such a horribly saturated market I wouldn't risk the capital outlay. There are better opportunities (lower risk/higher reward) elsewhere.

Posted: Thu Mar 08, 2007 7:59 am
by seodevhead
I do agree that this market is overly saturated, but then again, there are so many Hosting businesses out there with HORRIBLE websites that do not exemplify professionalism or trust, that I actually believe if you create an absolutely stunning and resourceful hosting website for your business, you can attract a customer base simply off it's professionalism. If you get inquiries about hosting, I would offer to give them a phone call, so you can put your knowledge and willingness to 'go that extra mile' for the customer on display. These things make a big difference in the hosting arena.

Posted: Thu Mar 08, 2007 8:55 am
by Chris Corbyn
onion2k wrote:
d11wtq wrote:It becomes incomprehensively expensive.
This. If you don't have an investment of £millions to hand then you won't be able to compete in the datacentre market. That said, a hosting company doesn't need it's own datacentre. Rent space in one.

What you will need instead is one of two things. Either awesome Unix admin skills so you can automate absolutely everything and run it all very cheaply, and thus compete in the low price market, or else you'll need fantastic sales skills to persuade people that you're worth spending a bit more money on.

Frankly it's such a horribly saturated market I wouldn't risk the capital outlay. There are better opportunities (lower risk/higher reward) elsewhere.
Yes, we're talking millions easily... it's not even borderline, it's huge.

Take a company like the one we host our project in ( http://www.internetf.co.uk/ ), they have so much money going into it and most of that goes towards hi-tech power systems and fire systems. It's also *a lot* of work. It's all very well having your automated scripts and your fancy UPS systems and your compressed gas fire safety... but if you don't maintain that to the highest level you're going to have to pay out a shed-load on failed SLA costs when you have a power failure, UPS failed to start, generators didn't work and customers all go offline for half a day! If there's one thing that's going to cause a huge problem, despite being such a trivial device, it's air-conditioning! When you have a room full of servers and the air-con stops working correctly it doesn't take very long at all before things start to fail.

I'm not saying don't try it, but it's a hell of a risk and takes years to build up. You will definitely need top-notch networking a UNIX skills too. It's no use providing customers with an internet line if they are going to conflict/interfere with each other...

As for renting space in a data center, well yes, of course you'll pass the hassle of maintaining the electrical/safety equipment over to somebody else, but you'll have higher overheads on the cost of keeping your servers there.

Have you worked in a data center before Hockey? Or if you haven't, have you maintained your own servers in one?

You'll also need man-power, and for that you'll have to pay salaries. Network admin jobs don't come cheap (well, ok, some companies don't pay very well..). It's not exactly the sort of thing you can take on single-handedly. Somebody has to physically be on-site 24/7.

So I guess the questions you need to ask yourself are:

Have you got millions of dollars, or can you borrow it?
Have you got the UNIX admin skills, or can you pay a handful of people who do?
Have you got networking skills?
Are you already familiar with data center environments and server management?
But most importantly...

Have you got any feasibility studies which show that you'll make money from it?

Posted: Thu Mar 08, 2007 10:19 am
by JayBird
Moved to The Enterprise

Posted: Thu Mar 08, 2007 11:32 am
by RobertGonzalez
I'd say start with looking into what it takes to start any business...

Building
Supplies
Employees
Legal preparation
Startup Capital
Marketing

Then throw in the core support items for the business that you are going in to...

Servers
Software
Licenses
Racks
Network gear
Fire systems
Electrical systems
Failovers
SANs and Backups

Add in maintenance and renewable expenses, and you got yourself an idea of what is needed to start. Then estimate the cost of all of that stuff, multiply by a factor of 1.5 and I'd say you have your target start up capital needs. Double it for the second year, add another 1.5 for the third year, and you can put something into a 3 year business plan.

Posted: Thu Mar 08, 2007 2:30 pm
by Kieran Huggins
As has been sort of mentioned many times on this thread, a data center is not the same thing as a hosting company.

Even Dreamhost, who is the second largest hosting company (I think) co-locates in 2 different data centers thast are not owned by them. They own the hardware and have 24/7 access, but the space, power/fire systems, etc.. is not their problem.

I would focus on what would distinguish you from the competition -- why would YOU choose your hosting over someone else's? Read the webhostingtalk forums for an idea of what problems people have and what they want from a host. I'd also recommend reading the dreamhost blog - it's full of interesting inside information, and they're privately profitable!

Once you have that plan worked out, buy some dedicated servers or space in a co-location facility and give it a whirl!

Posted: Thu Jun 28, 2007 1:23 pm
by inova
I agree with Kieran,

the most important thing is how you distinguish your products from others or you will end up being a freelancer with cheap projects

Posted: Fri Jun 29, 2007 1:35 am
by The Phoenix
Here's some thoughts (based on experience) on hosting companies, data centers, and so forth..

Like any business, the best growth is organic. That means starting with one customer, and when you get a second customer, expanding your business to accommodate them (or right before they join).

Putting together a "6-months from now, I'll probably be at.." business without the 6-month interim is how most businesses end up going under.

Now, as many have mentioned, hosting doesn't *require* maintaining your own data center. In fact, on a cost competitive basis, you are better off reselling existing hosting, or offering hosting from an existing data center. Low initial investment, and you build up a reliable customer base - which can move with you to a permanent facility that you maintain.

Assuming you can compete on features, price, and service against the *thousands* (not an exaggeration!) of hosting companies, you can then look at starting a data center.

Data centers are tough. You have three major concerns - Bandwidth, Power, and Heat.

Bandwidth is relatively simple (but expensive). You request a line from a provider, and you have to sign a 5+ year contract (generally). The contract gets you a nice pipe run to your location from their closest hop, and (usually) a CSU/DSU. You may have to pay for the 'last mile charge' to run the pipe to your location. Be prepared - the contract can be $15k plus over the lifetime, and the last mile charge could be as high as several thousand dollars.

Many of those costs are negotiable, depending on the carrier/location/moon phase.

Once you have bandwidth, now you need power. You can get as high as 200-300 watts per square foot in usage, and most simple commercial (read: Not warehouse) spaces cannot manage that load for a decent sized space. You also need the power to be 'clean', which means adding in load levelers, trip faults, and the like. Expensive again - easily in the $10-30k range.

Finally you need to get rid of that heat. Generally, the rule of thumb is "one ton per tile", and that can also end up being extremely expensive. For a measure you can relate to, a data center the size of one floor of a 3-bedroom house generally needs approximately a 6-7 ton cooling system (running constantly). Thats gonna require yet more energy.

And these ignore backups (power, cooling, data), security (physical, information, network), and design (raised flooring, humidity control, flooding, attack proofing). We should mention service in there too, as you'll need a rack monkey roughly 24x7 to compete with ANY of the established companies - thats full time pay, plus benefits.

Its a daunting, expensive, challenging, frustrating process. Worse, there are thousands of competitors at each stage:

1. Resell hosting (the main sellers can sell it for less, and offer better service)
2. Sell hosting (the main sellers can offer more features, better service, AND better uptime)
3. Small data center (the main sellers have more redundancy in their networks, better rack technology, and have solved the major problems - for less than you can)
Hockey wrote:My local service providers are the local phone company (DSL) and Shaw cable (national carrier - Cable).
Google "Canada backbone providers". LOTS of choices. DSL & Cable aren't terrible, but they usually won't get you the peering you need to be competitive. You want a Bell Canada, or similar. Someone with OC3+ trunks to feed from.

More than anything, it will come down to differentiation. If you can offer an incredibly (must be incredibly, not incrementally!) better set of features, service, and reliability, you MIGHT be able to make a dent in the market, as long as you scale organically.

Its not impossible, but its certainly improbable.