Data Centers optimize

SolutionBase: Steps to take to successfully move a datacenter

Moving a datacenter is no simple task. There are many things you have to prepare for and have ready before the big day. In this article, Brien Posey points out some of the most important steps you must take to make a successful move.

Whether you work for a large enterprise or a small company, you may one day find yourself having to move your company's datacenter to a larger facility. Needless to say, moving a datacenter is a huge job, even for a small company. The key to a successful move is planning, planning, and more planning. Unfortunately, every move is different, so there is no way that I can tell you everything that you need to know about moving a corporate datacenter. Having been through several moves, though, I want to take the opportunity to talk about some of the most important aspects of preparing for a move.

How long can you be down?

The first question you need to address in preparation for a move is how long you can afford to be down. In a large enterprise type environment, the typical answer is that no amount of downtime is acceptable. These types of environments typically have a backup datacenter that can handle operations while the primary datacenter is being moved.

Even if you have a backup datacenter available, you will still want to expedite the move; after all, there might not be a backup for the backup datacenter. If a catastrophe were to take the backup datacenter offline during the move, then both datacenters would be down. Of course, the chances of that happening are pretty slim, but when you are planning a move you should always plan for the worst-case scenario.

If you do have a backup datacenter at your disposal, you need to decide when you will transfer operations to the backup site and how you will go about verifying that everything is running there before pulling the plug on the primary site.

If you work for a smaller company without a backup datacenter, you must decide at what point the servers will be taken down. You must also decide how and when to let your customers know you will be offline for a while.

Prepare the new site ahead of time

In almost every case, your primary goal will be to complete the move and have the datacenter back online in its new location as quickly as possible. That being the case, you should try to gain access to the new facility a few weeks prior to the move. By doing so, you will be able to prepare the new facility so that on moving day all you have to do is to plug the servers in and bring everything back online.

So what types of things can you do to prepare the new facility? You could start by measuring the server room and creating a floor plan that shows where each piece of equipment or furniture will tentatively be placed.

Once you have a pretty good idea of where everything is going, you should check to make sure sufficient electricity is available. It could be that there are no plugs near one of your server racks. In that case, you will need to have an electrician come in and install some outlets for you.

Even if it appears that sufficient electricity is available, it's still probably a good idea to bring in an electrician to make sure that the circuits are connected to high enough amperage circuit breakers that your computers won't overload them.

Another thing you can do to prepare the facility ahead of time is to run network cables and crimp ends onto them. The same goes for any other types of cables that can be installed ahead of time. For example, if you use extra long cables with your KVM switches, then maybe you could install the KVM cables ahead of time.

If you have a sufficient budget, you might even be able to purchase brand new server racks for the new facility. As I'm sure you probably know, most server racks must be assembled and then bolted to the floor before servers can be installed in them. By purchasing new server racks, you could have the racks assembled and in place prior to moving day. That way, all you have to do is install the servers.

Take an inventory

It's a good idea to take an inventory of the hardware in your datacenter prior to the move. At a minimum, I recommend recording the make, model, serial number, and hardware configuration (memory, CPUs, hard disks, etc.) of each machine. Having an up-to-date and accurate hardware inventory will make it easier to replace any machines that are damaged or destroyed during the move. It will also speed the process of getting a settlement check from your insurance company in the event of a serious catastrophe.

Have a meeting of the minds

As I mentioned at the beginning of this article, moving your datacenter is a big deal, and you need to plan for the worst case scenario, even though a serious catastrophe is unlikely. This means that you must plan for the unthinkable. For example, what would you do if the moving truck containing all of your servers was involved in an accident and all of your servers were completely destroyed? The odds of something like that happening are slim, but the future of your company (and your job) depends on your planning for things like this.

I recommend having a meeting with your insurance agent and reviewing the coverage on your servers. Some of the questions that you will need to ask are:

  • If servers are destroyed, how long will it take to get a check to cover the cost of replacement servers?
  • Will my policy fully cover the cost to replace the servers, or will I get a check for the depreciated value?
  • Does my policy cover the loss of business that will occur as a result of excessive down time?
  • What deductibles would I be required to meet?

This is also a good time to give your insurance agent a copy of your datacenter inventory. Doing so insures your agent is aware of all of your hardware and gives your agent a chance to verify your policy is sufficient to cover any losses that might occur. While you are meeting with your insurance agent, you might also ask them if they would like to verify your list or if you need to provide photographs of each machine. This probably sounds a little odd, but my insurance agent required photographs of my hardware prior to insuring my network.

Once you have met with your insurance agent, it's a good idea to meet with your hardware vendor. The purpose of this meeting is to ensure that your hardware vendor will be able to get you through any moving related catastrophes.

Depending on your relationship with your hardware vendor, you might be able to negotiate a deal in which your hardware vendor is on call on moving day and the day after. There are a couple of reasons why it's important to negotiate a deal like this.

First, servers just do not like to be moved. I have been through several datacenter moves, and I have never had one go completely smoothly. It always seems as though I have to replace a couple of hard drives -- or maybe a power supply or two -- after a move.

Second, it's important to have your hardware vendor available to you because your ultimate goal is to minimize the impact of the move on your business. As such, moves often happen on the weekends; typically late at night. Under normal circumstances, it would be difficult, if not impossible, to find a computer store open at 4 AM on a Sunday. You really don't want to have to wait until Monday morning to be able to replace any damaged components, so your hardware vendor needs to be on call to ensure that any necessary components can be purchased immediately.

Once you have gotten your hardware vendor to agree to be available to you during the move, you need to address the issue of potential serious catastrophes. For example, you need to know how long it would take your hardware vendor to get you replacement machines if all of your servers were destroyed during the move. Since it would likely take your insurance agent a few days to get you a settlement check, you should find out if there is any way that replacement machines could be ordered (and possibly even delivered to you) before the insurance check arrives. Your hardware vendor might be reluctant to agree to such an arrangement, but they might change their mind if you remind them of the amount of money you spend with them each year and how much they would profit if you replace all of your servers.

Backup everything

Another critical task related to moving is that of performing a full, system state backup of each machine prior to shutting it down. Remember that your servers will be in a precarious position during the move, so these backups are basically your lifeline. You must therefore take the backups very seriously.

At the very least, you will want to verify the backups; but ideally, you should try restoring them to another machine to make sure that there are no problems. After a backup is made, it should be stored in a cool, dry, and secure place until after the move.

You have probably heard the old saying: "Don't put all of your eggs in one basket." Well, this saying applies to this situation. On moving day, don't put your backups on the same truck as your servers. Otherwise, the backup tapes could be destroyed right along with the servers if the truck was involved in a serious accident. Instead, transport the tapes separately.

When moving the backup tapes, be especially aware of the temperature in the vehicle that is being used to transport the tapes. If you are moving in the winter, you will want to ensure that the vehicle's heater has warmed up before you start loading the tapes; cold weather can wreak havoc on backup tapes. Likewise, if you are moving in the summer, don't leave the tapes in a hot car.

Prioritize which systems should be brought back online first

One of the most important things you should do is prioritize the order in which the servers should be brought back online once they arrive in the new location. In any organization, some servers are more important than others. Your goal should, therefore, be to bring the highest-priority servers online first, so as to minimize the downtime of mission critical applications. If you don't have a backup datacenter, you might even be able to orchestrate a plan in which the highest-priority servers are brought online in the new facility before the lower-priority servers are even moved.

Prioritizing your servers helps minimize the downtime for your mission critical applications, but it also helps you to spot any critical problems quickly. Imagine, for example, that you decided to move everything at once, and then bring all of the servers online once everything had been moved and hooked up. If one of your most critical servers got damaged during the move, you would end up being down for longer than you intended. If you had focused on moving and bringing up the most critical servers first, you would have found out about the problem more quickly and could have start working on it right away.

10 comments
ljansch
ljansch

The processes involved in moving a datacenter should come right out of the center's tested and mature Business Continuity Plan. A relocation is nothing more than a controlled disaster with little, if any, equipment damage. BTW, backing everything up to tape without testing the procedure with a recovery from those tapes is very risky. - Larry

doogal123
doogal123

When moving servers, you may find that the mfg. warranty will be void if YOU move the servers. Sun HP/Compaq will apparently void your warranty if they do not do the movement of warranty covered servers themselves. They have to be the ones (and you pay them) doing the server tear down, packing, unpacking, reracking,and power up of the servers at the new location. Just learned this as a contractor on a data center move.

curt.watson
curt.watson

We had a successful Data Center move 2 years ago in which we spun up a new facility with ~200 servers. Here are the keys: 1. Buy new servers and network infrastructure. There are going to be one-off servers that simply have to be moved (can't find the source code or recreate), but it's way too risky to move the equipment. 2. Set up a huge data pipe from the old facility to the new one. In our case, we were able to migrate 1500 Exchange mailboxes over a 3-month period prior to our final migration because it ultimately didn't matter to the user where their mailbox resided.

slands10
slands10

Remember to plan for areas outside the dataroom when killing the power to the building - when we changed to a new power supply and generator / UPS, we forgot about the fridge freezer upstairs which started to melt, luckily we realised in time :-0

MakeItSo
MakeItSo

I'm IT Manager at a High School and in December we will be remodelling the centre of our star network, 2 full height server/switch cabinets, pabx, UPSs etc currently in 3 different places in a block also including 12 classrooms. The current plan is to build a 'datacentre' first then move the cabinets and pabx etc - about a 10 metre move - into it. Then the builders will be remodelling the rest of the block. We also have to move 140 PCs somewhere else then bring them back! The article couldn't have come at a better time and it has some great suggestions. We've just atarted planning now and have a tentative floor plan for our datacentre so I guess we're on the right track. I'm concerned about dust, power loss, fibre being damaged etc. At least it's not so critical for a school network to be running over the summer break so would it be better to switch it all off and move it away then bring it all back after the work is finished? Any other suggestions?

psacco
psacco

Additionally, pre-migration is the perfect time to virtualize your server platforms. It not only will allow you to replace fewer boxes, but the migration of data is far faster, secure, and safer than a physical move. Pete Sacco

Dr Dij
Dr Dij

APC, a vendor of modular racks, UPSs has free courses on how to minimize costs: click on register new user; there are 50 some costs. http://lms.globalknowledge.com/ilearn/en/learner/jsp/clients/APC/customer/login.jsp They are reasonably vendor neutral. They plug their modular UPS and racks. What was mind blowing tho is that there are many misconceptions about data centers: you should have raised floors; false they are a relic from mainframe ages and have MANY SERIOUS disadvantages and reduce total uptime. You should have a huge UPS; false, they are terribly inefficient You should oversize greatly: false; with modular build out you don't need to. Also learned many other things such as electricity being one of greatest portions of ongoing expenses and many others.

rich_soby
rich_soby

Remeber Murphy - it sounds good to turn it off, move it and turn it on, but nothing is that simple - the gremlins will be there - Also, make sure power and cooling is adequate - We did this and it still got messed up

psacco
psacco

Having been a data center designer & engineer for many a site, I must say that I disagree with your assessment not to include a raised floor. PTS has installed hundreds of APC's InfraStruXure series UPS, CRAC, and Racks and I must say we always recommend using a raised floor even when we are utilizng in-row or overhead cooling. There is just no better location for refrigerant and water lines as well as large AC power feeds in conduit for remotely mounted PDUs. I agree with your points on right-sizing both power and cooling for operating efficiency as well as your view on scalability. Pete Sacco

dchow
dchow

Thanks man, I'm a newbie college student still in entry-level administration. That link clears up alot!

Editor's Picks

IT Buying Cycle

Learn more