In general, how soon can you justify a datacenter? By datacenter, I'm not talking anything large, I just mean getting a T1 or something with your servers in a back room of your office. Is this a wise choice to persue once you get an office, or have you found co-location to work well for a while?
Colocation is better than "just a T1 in the back room".
The key to having services worth purchasing, as you know, goes well beyond the server itself. You'll need MULTIPLE high speed bandwidth pipes connected to it, a solid power supply with backup power provisions, as well as security etc etc etc.... The server itself may fit into a desk drawer and there's nothing stopping someone from hooking a T1 to it. However, if everything else that should be surrounding that server is missing there's bound to be a problem along the way..
For this reason I recommend researching "now" everything that is required for an efficient data center. It doesn't need to be big, it just needs to have all of the required functionalities. Once you have mapped it out and figured out what the total costs will be, you'll then be able to successfully judge when your company is going to be able to sustain such a venture.
It's certainly worth it in my not so humble opinion but until you are ready for it... all things that go with it...I'd recommend colocation...
Unless you really ready for all of the 24x7 needs, you much better getting a rack (or multiple) racks at a reputable data center. Data Centers are already staffed 24x7 and can be contracted to perform tasks for you at odd hours of the day. We have a break/fix contract in place that activates when we do not respond to a call within 15 minutes. Our customers are never at the mercy of our availability.
I have connections in a local telco facility where I could be a couple racks and associated BW much cheaper, but the place isn’t staffed. Looking at the cost of providing adequate staffing, I’m better off with multiple racks where I am.
Thanks for the responses. I understand what's involved, as well as the costs, but I guess what I'm asking is, based on your experience, how many servers would justify the costs involved? My point about the office was, if you already have need of an office, then the cost of the building isn't so much a factor.
Obviously, there are a number of benefits to running your own datacenter. You have more control over your servers and network, and its readily expandable, much more so than having to order another dedicated server, or ship one of your servers out there as there is demand.
I agree about the added control being a huge benefit. I couldn't imagine going back to colocation after gaining the abilities that we have by operating our own but deciding when the time is right is rather complicated as there is no set equation.
You have to look at the amount of revenue your servers are bringing in and compare it to the expense of having the data center and then decide whether or not it makes sense. An "office" does not generally have everything a "data center" requires. Our fire system requires our DC to be AIR TIGHT which is something most office space is not. Our Backup power includes both the UPS and the gas powered generator to ensure the servers do not "blip" during a power outage. In an office environment a quick "blip" usually isn't as big of a deal so there may be a second or two of "lights out" during a power failure. We also had to have the power company come into our location and add an additional, much larger, transformer to cover our power requirements. The normal "office space" power was not enough to cover the needs of all of our equipment.
There are simply a ton of factors to consider when making the decision. How much you charge for each account and how many accounts you can fit on a server are going to directly affect how much money you have to afford the extra requirements needed to operate your own DC -- and as was noted above...never forget the cost of manpower
Like Deb said you have to weigh the costs vs. the income you bring in, and projected income with your own DC. If you have a plan to do better in the long run, then it might be worthwhile. Besides the costs of the network, office, upgrades, etc... 24/7 staffing becomes an issue, a spare parts inventory for failed hardware, routers, switches, etc..., security of the network itself, plans for handling DDOS's, and the list goes on.
For us, we have around 50 servers, and would not even think about our own datacenter. Even colo is not attractive. We pay a good amount (six figures) per year to our DC for the dedicated servers, and they do a great job at maintaining the hardware, network, and 24/7 manning that we do not want to do ourselves. It gives us peace of mind, saves us on employee costs, and allows us to sleep well at night without the 3 AM pages.
In 1999, prior to the DOT COM bust, we hired an architect firm to develop the plans for a reasonable size data center with an estimated location in either Fairfax VA or nearby.
The result of the planning was that building and land costs alone would be close to $1,000,000.
In 2000, I had the opportunity to hear Sean Brophy, Vice President of NTT/Verio, speak about data centers; he shared that most modern data centers (at leas the type NTT/Verio build) cost $600 per square foot to build, and $300 per square foot to maintain.
Deb from FutureQuest is very knowledgeable; most office environments are night and day different from a data center environment. The differences are often vast enough that from a marketing and advertising point, those who rent or co-locate in a real data center have an edge over the “in your house, garage, office, etc.” environment.
For example, when we first started hosting in November 1996, I was approached by a local Berks County business person who started their own “in their garage” data center.
They had two T1’s coming into their garage with computers here and there (though some what organized; they didn’t use racks from what I can remember).
Barry, the owner, offered to give me a 386 PC to host our Web sites (remember this was 1996 – granted, 486’s were available at the time… not sure when Pentium’s came out in retrospec).
While we co-locate now, we were a reseller of RapidSite (now, an NTT/Verio company) in 1996.
Barry could not understand why I would not take him up on his offer. He offered to match or beat their rates; and I would have my own server (free).
Our first client was an on-line bank. Now, think of it from the bank’s perspective.
Hosted on Silicon Graphics mini-computers running secure IRIX in a real data center boasting six (I think it was six at the time) DS-3’s with routing done by BGP4 OR hosting on a 386 within a garage off two DS-1’s.
Which one do you think the bank would pick?
Now, fast forward to 2004. If your customer base would care less that you have one or two T1’s (probably not running BGP4) in a garage (or similar) type set up with little security (in comparison), close to zero environmental controls (again in comparison), and so on, then go for it.
Originally posted by dynamicnet Barry, the owner, offered to give me a 386 PC to host our Web sites (remember this was 1996 ï¿½ granted, 486ï¿½s were available at the timeï¿½ not sure when Pentiumï¿½s came out in retrospec).
The Pentium was released in March 1993. I remember particularly because we scored several brand new ones for a project I was working on - they were great for Doom