Posts Tagged ‘infrastructure’


You need POWER:

  • Clean and reliable
  • Batteries. Need to support your entire infrastructure for at least 150% of the time it takes for your generator to come online so that if there is a problem with the generators coming up you have some wiggle room.
  • A generator capable of powering you entire infrastructure with at least 48 hours of fuel on site and contracts for delivery with at least two companies to sustain you after that.
  • Scripts to properly shutdown everything if your generators give out and you are on batteries.

You need NETWORK

  • Your one 9mb pipe from a single ISP isn’t enough. You need more bandwidth, and redundant ISPs.
  • On the topic of redundant ISPs you need to make sure that both the back end and last mile are completely different. If there is a manhole fire the burns through a fiber line, and that one line was were all of your ISPs were using as their back end network, you have a problem.
  • You are going to need a lot more bandwidth than 9mb.
  • You are going to need to completely understand how WAN routing works and be able to have someone who can diagnose problems quickly as downtime is not tolerated at all.
  • You are going to need a relationship with ARIN to buy your IP blocks.
  • You are going to need to make sure that not only your ISPs are redundant, but so is your entire core network. Having one end of row switch to pass traffic to every rack is fine, but your entire core needs to be fully redundant. AND YOU NEED TO TEST THAT. A perfect example is the RIM outage where they thought they had redundancy, however there were issues with the routing and the secondary core switch didn’t fail over properly.
  • You are going to need to have IDS and IPS setup to detect and prevent attacks when possible, as well as DDoS mitigation.

You need COOLING

  • You need to be able to keep your DC at an even 70 degrees F whether it is 30 outside or 107. If you use a standard portable AC unit, that is fine, if you use a built in commercial style unit that is even better. No matter what you do you need to make sure that your cooling is tied into your batteries and generator as well. You also need to make sure that you have a secondary cooling solution if one fails.
  • You need an exhaust of some kind to remove all the warm air.
  • You need better monitoring for your cooling than just a wall thermostat. Having at least 3 temp sensors in each rack is ideal. This needs to be network monitored and alarms need to be set.


  • Making sure no one except authorized users are in the DC is critical. Keyfobs are good, biometrics are better. A log of access times (both in and out) should be kept.
  • Full time video monitoring with all angles covered.
  • Making sure your racks have individual keys and not one generic lock (HP racks used to be really bad about this).


  • FM200 or Halon. If you put water in the DC that will soak everything. If you are Colo-ing equipment your insurance may not pay to replace that and then you are out money.


  • Doesn’t matter what you use, you need real time monitoring of all servers, switches, routers, firewalls, traffic flow, temperatures, power usage, the works. Anything that belongs to you needs to be monitored for your 99.9% uptime guarantee.

You need BACKUP

  • You need backups of your configs for all your stuff.
  • If you are hosting servers on your own equipment you need to back that up nightly
  • You should offer backup services to customers to be competitive.
  • You need to test your backups every once in a while.


I’m not going to bullet this one, but you need something that makes you different than everyone else. I have 5 CoLo DCs in my town, not to mention The Planet (Layer 8) / Liquid Web / AWS / EC2 / and many more I am forgetting. What makes me want to use you to host my servers instead of giving them to a bigger name that is trusted and proven. If it is because you are cheap? Well if you are cheap is that because you are new, or because you don’t have the same investments in infrastructure that the big guys have and are therefore more likely to give me problems? Is it because you are local? Well if you are local who else are you fighting locally and what do they do wrong that you do right? You really need to come up with a business plan that makes people want to be with you. There are a TON of people in this business who have been in it for longer than you, who started out with more experience than you and failed to get a hosting / colo company off the ground so what makes you think you can succeed where they failed other than the fact that you may be a greater fool (not a personal attack, just trying to get you to take a long hard look at yourself before you invest a half million dollars into a pipe dream).

Speaking of which….


Seriously this stuff costs a lot. Right now my bandwidth costs in my DC alone are about $2000 a month, servers aren’t cheap (doesn’t matter if you go supermicro barebones and have spares on hand, or buy dell / hp and have 4 hour restore contracts), networking equipment isn’t cheap, and power isn’t cheap. Not to mention you are going to have to pay people to keep this running for you (which isn’t cheap). To give you an idea, my company offers a small hosted exchange, hosted spam, and hosted email encryption service (we have about 2500 accounts between the 3). Our equipment (2 HP DL360 G6, 2 HP MSA P2000s, 2 HP 2910als, 2 Sonicwall NSA 4500s, supermicro barebones storage server, 2 HP DL 180s, and a few other odds and ends) cost us just shy of $100,000 in hardware and support contracts, not to mention the other $150,000+ we have paid over the last 5 years in microsoft and other licensing. So in our first year in business we probably spent about $200,000 just to fill the rack, power it, and get external customers access to it. Again not trying to throw out big numbers to scare you, just give you the realities that this stuff costs money, and doing it “on the cheap” will bite you in the ass hard down the road.