Data Centers

10 mistakes you might be making with your data center

Certain common pitfalls and planning mistakes can cancel out data center benefits, like enhanced reliability and cost savings. Here's what to watch out for.

If you're a small to midsize business, you may be thinking about migrating your servers to a data center -- either by renting space in an existing data center or building your own. Regardless of which route you take, it is imperative that you avoid falling into the traps that have caught many before you.

Thankfully, because others have made those errors, you can avoid them all together. But how do you know what to avoid? Here are 10 common pitfalls to watch out for.

1: Inadequate virtualization

Let's face it: If you're not virtualizing, you're already behind. And if you're in a data center and you're not virtualizing, you're not only behind, you're spending more money than necessary. Why have dedicated hardware for every task when you can have a single machine hosting virtual servers? Not only should this save you money, it will save space and will (ideally) become a far more reliable environment.

2: Limited expandability

If you set up a data center without thinking of expansion, you're limiting yourself. This means building racks to accommodate growth, putting plenty of RAM and drive space in servers, and even designating enough space for your data center. As a rule of thumb, consider what you need and double it. That way you'll know for sure you have enough room for expansion.

3: Lack of cloud resources

There are plenty of cloud naysayers. But the truth is, the cloud is a flexible resource that every midsize business should leverage. You can create your own cloud with Ubuntu Server Edition or purchase cloud space from Amazon. Either way, you set yourself up for a far more flexible data center experience. Even if you use the cloud only for storage, at least you can constantly expand your storage without having to constantly add new hardware.

4: Poorly managed servers

We've all either seen the results of poorly managed servers or we've managed servers poorly ourselves. When this happens in a data center, things can go sideways very fast. Be sure to keep your servers up to date, backed up, and plugged into battery backups -- and check on them daily. Just because a server is sitting in a data center, that doesn't mean it's immune to problems.

5: Badly designed and cramped racks

How many times have you had to work on a rack and couldn't weave your arm through the cables or even get to the patch panels? Racks have to be designed so they're easy to work on (from front and rear). Patch panels must be clearly (and correctly) labeled, and cables need to be properly (and intelligently) tied. Do not make the mistake of tossing everything together just to get it up and running. Take your time and design.

6: Underpowered facility

If you don't have enough power for your data center, you will "blow fuses" (or worse). And what happens when you have X number of servers and only Y number of electrical circuits? You do not want to deal with this special flavor of insanity. When you design (or select) your data center, be sure you do so with an eye to power. Go for too much rather than just enough or too little.

7: Weak security

This shouldn't even have to be addressed -- but it does. Many people assume that because they are going with a data center, security is no longer a concern. Don't let that tragic assumption come back to bite you. No matter where your servers are, always make security a priority. No data center can guarantee with 100% assurance that your data is safe. Why not treat your servers as if they were sitting directly on your own pipe and have that extra layer of protection against threats?

8: Poorly designed remote access

You have to gain remote access to your servers, both from within the company and without. Make sure you have more than one method of getting in -- and make sure those methods are secure. If you rent space and it offers a control panel-type remote gateway, it's okay to make use of it. But don't rely on it.

9: Shortsighted budgeting

If you budget your data center in such a way that it will power your current load, you are shooting yourself in the foot. When you budget, make sure to have an idea of what your needs will be in the next five or 10 years. Otherwise, expanding could be more of a challenge than you expect. Again, buy more, not just enough or less. Always assume you will grow and go with projected numbers, not current numbers.

10: Failure to alert clients of maintenance or downtimes

If you host clients on your data center (this can be paid clients or even departments within your company), make sure you alert them when downtimes will occur and for how long. If a problem arises and the server goes down for some reason, alert the users that you are aware of the issue and are working to resolve it. Don't just take down a server without alerting -- you will not only cause data loss, you will quickly sour a relationship that your bottom line might depend upon.

Do it right

Data centers have been popular for a long time for good reason. They enable a business to expand, and they're cost effective and reliable. If you are about to purchase data center space or build your own, make sure you don't fall into these traps and wind up in a state of chaos or major rebuild/redesign.

About

Jack Wallen is an award-winning writer for TechRepublic and Linux.com. He’s an avid promoter of open source and the voice of The Android Expert. For more news about Jack Wallen, visit his website getjackd.net.

12 comments
avindia
avindia

Based on my recent experience with Managed Hosting (private Cloud provided by Third Party) is alway expensive compared to maintain own DataCenter.The only advantage you get is moving away from overheads like realestate,power and upgrade cost. You should not look at cloud for Cost saving unless you have applications with less duty cycles (server transations happens once in day or occasionally). You will pay only for actual power usagae and server utilisation.

orovan
orovan

Just wanted to relay my own experience, don't mean to stir up any love-hate discussions about AWS. I support IT infrastructures for small businesses and have been working the numbers with every new deployment. Anymore, the cost of doing it on your own with your own hardware and physical location is by far higher than building it on AWS. I run most of my servers in spot instance and even in a month like last month that was expensive for spot instances, the monthly running cost was still below cost of reserved instances. When you total running, storage and IOs cost for a three year span you will come up with less of what you would have spent on your own data center. And all that is paid in monthly payments for resources used as opposed to lump sum purchase or worse, interest bearing payments for your own physical infrastructure. If you can get away with placing your infrastructure a WAN link away from your physical location then don't be wasting your time and money with building your own data center. AWS had a bad outage last month and of course it pissed me off but I stick with them because their prices are unbeatable and their services are constantly improving with new ones coming out every year. I did not mean to make this sounds like an AWS infomartial, just to relay my own experience and considerations with regard to the topic this article covered.

alexisgarcia72
alexisgarcia72

Another issue I would love to add is a proper shutdown procedure. In today datacenter enviroments, is pretty common to have a mix of routers, firewalls, NAS, SAN, Vmware servers, backup servers, tape drives, email servers, database servers, UPS's, switches, Intrusion detection devices, Proxies, WAN accelerators, etc. Some small datacenters can have all this equipment but lack of power generation equipment. In a case of extended power failure, power maintenance and other issues where you will have no power for several hours (or days) in the datacenter, You will need to shutdown all your equipment in the correct order and with the right procedure. For example 1- never shutdown your domain controllers before shutdown Exchange first, 2- make full backup (if this is an scheduled power down) of all servers 3- turn off servers in the right order: for example if you have a Blackberry server (BES), turn it off before Exchange 4- turn off SAN devices at the end of the servers turn off routine 5- have proper docummentation of the power off routines for special devices like SAN, virtual disks and other devices in use by Virtual infrastructure (you definiitely do not want to turn off the SAN where the ESX servers have the data) 6- have docummentation of all ip addresses, login and passwords, 7- have a copy of the clients you will need to power on / power off your units (ie: vmware clients, san clients, etc) 8- consider to have a laptop with extended battery time in your site or office 9- list of all IT contacts, providers, etc 10- notify users about power disruption, in maintenances or emergencies 11- some ISP can cut your E1 links if your device are turned off fore some x time. I normally turn off all equipment but I have a big UPS able to provide enough juice to the router for about 8 days. In this way I avoid our ISP to cut our ring and have problems when power is back Hope this works for you :)

doncrawley
doncrawley

Although events like Hurricane Sandy are rare, adequate disaster preparedness can help minimize the impact of such events. Backup generators can't do their job when they're under water, nor, for that matter, can a data center function under such conditions. Natural disasters such as flooding, tornadoes, earthquakes, and volcanoes can render even the most robustly-built data center useless. Have a separately-located failover site.

sudheesh.pb
sudheesh.pb

11th. Poorly designed network infrastructure 12th. Network Security Issues

neil.postlethwaite
neil.postlethwaite

Sorry, most of the above are the bleeding obvious, and if you make them, you are obviously in the wrong job.

robo_dev
robo_dev

is a bad thing? :) I think perhaps what you are hinting at is actually TESTING your DR or BCP scenario. Lots of the things you mention sound like something learned the hard way. It's much better to test all these things on a sunny Saturday versus when you're knee deep in flood water.

orovan
orovan

Don Crawley - I used your Linux server book about two years ago and it was of great help. I came a long way since but it sure got me on a good start. Just wanted to say thanks!

rduncan
rduncan

they are obvious but that's achems razor all things being equal the most likely explanation is the correct one. it is a good list of mistakes that IT departments keep on making - as for being in the wrong job- you could make a list like this of any industry, it just highlights the big issue, stick to the basics and have a strong foundation, embrace change and don't be technophobic

robo_dev
robo_dev

I agree that most of these are blindingly obvious, and may perhaps be to educate someone with little knowledge of how a proper data center works about what is what? There are some businesses who have what I would call a 'pseudo data center' that is some cobbled-together office space with a strong lock on the door. I would think perhaps that would apply to them.

eds
eds

It is surprising how easy it is to miss 'The Bleeding Obvious'. Perhaps the fact that a hazard is familiar makes us less wary than is prudent.

alexisgarcia72
alexisgarcia72

Actually I don't learned this in the hard way, instead, I documment everything and I experience myself some medium size datacenters where I see 1 - 2 hours shutdown process is required. Last year in a power maintenance we expect to be without power for 4 hours and after 10 hours the energy is not comming back, so we plan for everything before happens. regards

Editor's Picks