Planning a Datacenter Migration

Moving your IT operation to a new datacenter is an enormous and intimidating proposition. Whether you are moving from one in-house location to another or plan on outsourcing your datacenter to a third-party hosting facility, planning and preparation are key to a smooth and successful transition.

This article will help you to understand the process and what to expect along the way. We'll begin by getting a full picture of your existing infrastructure, plan for future growth, choose an appropriate upgrade path, and ultimately pull it all together in a brand new location. When you break it all down, you end up with several small and manageable steps.  This "baby steps" method allows you to effectively manage your move and efficiently communicate your needs and plans with vendors, contractors and members within your own organization.

This article covers a few of the things you should be prepared for while moving your in-house datacenter or server room to a third-party collocation facility. While many of the topics discussed here may seem rudimentary to some, others may find this information very enlightening.

Inventory Your Assets

Network Diagrams

Before you can even begin to think about moving your gear to a new location, you must know exactly what you'll be moving. If you've put off updating your network diagrams, now is a good time to pull out those old Visio or Dia documents and bring them up to date. Even a simple spreadsheet can be a very effective way of visualizing a rack elevation. Using a spreadsheet in this way also makes it easy to track make, model and serial number, current draw and IP addressing. If you're planning a lateral "forklift" move from one location to another, this drawing will serve as the blueprint for you and your contractors while they "rack and stack" your gear in its new space.

In addition to documenting your local networks, put together a list of WAN and MAN circuits, T1s, DS3, PRIs, etc., that will need to be moved to the datacenter. Most telcos will allow you to "swing" one end of a circuit without paying a contract penalty. You'll pay a one-time engineering fee and your monthly rate will be adjusted if channel mileage increases. In some cases, you may find it necessary to order new circuits and cancel or re-use older circuits for other purposes. These are things you should keep in mind during the early planning stage.

Naming and Labeling

An important part of documenting your network is naming and labeling. It's a very simple thing to do, but I constantly find networks that are troubled by inaccurate or completely absent labeling. While putting together your rack elevation drawings, take the time to apply missing labels and verify machine names and the services each server provides. If you don't have a handheld label printer, get one! It's also worthwhile to invest in a label printer than can generate wire tags.

A sensible naming convention goes hand-in-hand with your labeling efforts. If you have multiple domain names, internal and external networks, redundant firewalls and load balancers, make sure you properly identify each device and get them labeled. When your primary firewall croaks, the last thing you want to do is powercycle the "secondary" only to discover that it was really the "primary" bearing an inaccurate label. Come up with a good naming convention and use it consistently, from DNS to physical labeling to asset tracking and accounting. One box; one name.

Space and Power

While it may seem obvious, understanding your space and power requirements can be a very time-consuming part of migration planning. If you're lucky enough to have intelligent PDUs or power strips in place, you're way ahead of most folks and probably have a pretty good grip on your energy requirements. More often, it's a loose estimation based on the number of power circuits feeding your datacenter or server room. If you're migrating to a space that offers redundant isolated power circuits, there is one very important thing to keep in mind with respect to high-availability power: Circuit loading and failover surge. A pair of 20-amp utility circuits feeding a stack of servers with dual power supplies costs quite a bit in a datacenter environment. You may be tempted to load up both circuits to capacity. Don't do it! If one of your power circuits should fail, the resultant surge imposed on the remaining circuit could trip a breaker.  You could find yourself with a dark rack, even after you've invested in dual power supplies and utility feeds. Each 20-amp circuit should be loaded only to about half its capacity. This means you'll want to avoid going over 10 amps on each circuit. To really play it safe, keep it down to about 7 or 8 amps per circuit.

Assessing Current and Future Needs

A Curious Situation

With servers getting smaller and more energy efficient, it has become a little bit tricky to plan for space and power requirements. One rack of gear with two 20-amp power feeds can house hundreds of times more computing power today than it could just a few years ago. Moreover, server virtualization throws another curve into the mix, since many of the services provided by dedicated machines can very well be virtualized and aggregated onto a single physical device.

Developing a Growth Plan

At this point, you have an accurate inventory of your servers and the services they provide and you should be looking at what you're going to do with them. Will they be allowed to move to the new datacenter? Do you plan on replacing them with new gear and migrate only services and data? Are these machines candidates for virtualization? These three questions play heavy on your growth plan because they likely will have a direct impact on your space and power requirements going forward.

Bandwidth and computing capacity are important considerations for any growth plan and play an important part of planning your move to a collocation facility. Bandwidth in a hosting facility is provided on a metered burst basis, whereas bandwidth in your office via T1, DS3 or other form of broadband is usually flat-rate. For this reason, it is critical to know your bandwidth requirements at contract time as overages can be very expensive, sometimes by a factor of 10 or more!

The services you plan to provide from the datacenter will make a difference in the type and size of internet connection you will need. If your servers will provide services to the internet at large, your network and bandwidth requirements probably will be different than if you are providing internal services to your local users over a dedicated point-to-point circuit. The flexibility you can achieve in a collocation facility makes both scenarios possible. Here in Orange County, the availability of high bandwidth point-to-point metropolitan area networks makes it very affordable to connect your office to the datacenter. This makes it possible to have LAN-speed connections from the desktop directly to your collocated directory services, email and file servers, and ultimately the Internet.

Know Your Business

In order to effectively plan IT growth, it is important to know where your company is headed. Unless you're in an IT-related industry, it's quite possible that your VPs and directors may discount the importance of technology in their planning and growth strategy. If this sounds familiar, it's critical that you make some friends at the top and get your concerns to someone who will listen. Make your management aware of weaknesses in your IT infrastructure and provide them with the information they need to get things moving in the right direction.

Where is Your Industry Headed?

Your CEO keeps tabs on the competition; the CTO or IT director should also know what the competitors are doing. Are they going it alone or do they have outside help? A little bit of legwork and research may help with your own growth plan. And it may even improve your chances of getting the budget approval you need to procure the technology upgrades that will keep your company operating efficiently and reliably.

Where is the IT Industry Headed?

Changes happen fast in this sector. Buying forward is a good way to optimize long-term value while minimizing the frequency of upgrades and the stomach-churning budget negotiations that accompany these expenses. Buy today the computers you need next year and they'll serve you well for a long time. Buying cheap commodity hardware on its way out pretty much ensures you'll be doing it all over again very soon. Early adopters often will pay a premium for new technology, but joining the game when the second generation hits the streets usually brings great value. Watch for new technologies and consider how these technologies could optimize your infrastructure by reducing energy and space requirements and minimizing administrative and support costs.

Datacenter Analysis

Finding a Collocation Partner

When searching for a collocation service provider, you will face three major considerations:

• Location. Since you will need to visit the site from time to time, you'll want to find a facility that is within reasonable drive time from your office.
• Services. You will need to work with their sales team and sales engineers to determine if this facility offers the services you require.
• Price. You'll likely find that the quoted prices for service range widely from one collocation service provider to another. The trick here is to get quotes from as many providers as possible, then figure out where they differ. Some will charge excessive "per-port fees" and monthly "maintenance" fees on cross-connects, while others may charge more for power circuits.

Amenities and conveniences are not as important as the operational capabilities of a hosting facility, but they certainly can make your visit more comfortable. Some facilities offer free use of conference rooms equipped with internet connections and projection screens, internet kiosks, crash carts and hand tools. Some even offer complimentary snacks, pretty good coffee and shower facilities. This may not seem important at first, but it could save the day if you need to make an unexpected visit to the datacenter on a Sunday afternoon or when you spend an all-nighter in the datacenter and need to be back in the office at 8 a.m. to deliver a progress report.

Network Availability and Peering

Topping the list of services offered by a collocation facility, of course, is the availability of network services and peering. Most modern hosting providers offer aggregate network connections which will be announced out to their own network service providers. Depending on your needs, this may be completely adequate. If you provide internet-based services and simply cannot tolerate downtime, you may need to take connections from multiple providers and announce your own networks via BGP. Discuss this with your sales engineer before choosing a collocation provider. Some providers charge management fees or costly one-time setup fees for customers who wish to speak BGP. Other providers don't even offer BGP or third-party internet connections in their datacenters. It's important to know your needs and what you expect from a service provider before you sign a contract.

Front-Office Backhaul and VPN

When searching for collocation providers, take the time to discuss with them the types of secondary services they can provide. If you're putting together a network with multiple sites, seek out a global provider that offers "on network" VPN or tunneled or switched point-to-point connections on their own backbone. In many cases this can save money and generally provides lower latency and better network continuity than building your own Internet-based routed VPN. Some providers can even extend their on-network VPN services via WAN or MAN links to the exact location you need.

Electrical and HVAC Facilities

Any reputable collocation facility will take great pride in their electrical services and cooling/ventilation systems and will showcase their gear and highlight its capabilities during your initial visit. If they're not willing to show you their UPS room, switch gear or generators, keep shopping. High-end facilities will generally have two utility feeds, multiple backup generators, and isolated dual power all the way down to the static transfer switch that sits just in front of your breaker panel. Smaller, low-costs providers may have a single generator with automatic switch gear and rack-mount UPS gear. In smaller markets, you may not have much choice when it comes to collocation providers and the level of redundancy they provide.

Physical Security and Operational Staffing

Many of the larger internet datacenter operators today are reducing operational staffing and centralizing much of the provisioning and monitoring duties that were previously handled on a per-site basis. Again, it's important to discuss this with your sales rep before signing a long-term contract. It may make sense for a large operator to collapse these services to a central location, but in most cases, we have found that local service usually suffers as a result. If this concerns you, and you don't believe you will benefit from global services offered by one of these larger companies, you may find one of the smaller providers better suited to your needs. In the current environment, smaller providers likely will provide more personal attention and may be better suited to handle the needs of small to medium customers. Enterprise customers with a large internal IT staff may not find this as important.

Physical security is part of the deal when you buy services from a collocation service provider. At the very least, they should provide 24-hour onsite security personnel. Many sites have security personnel on staff, others may contract with third-party security providers. In general, we try to avoid "lights out" facilities that do not have round-the-clock onsite security.

Lateral Move or Upgrade?

This is Your Chance to "Do it Right"

With all the planning and preparation that goes into a project of this scale, you may want to consider this your golden opportunity to fix things that have caused problems for you in the past. Replace troublesome servers, buy that new SAN you've been dreaming about, upgrade switches and routers, segregate services, and get a grip on your cabling. You've planned for this down time; you may as well utilize it to the maximum extent possible.

"Forklift" Migration

The simple concept of picking up everything and moving it to a new location is sometimes called a "forklift migration." Sometimes it is the easiest way to get from point A to point B, but there is still a great deal of planning and preparation that goes into this kind of move. Since you may have different teams working at each end of the move, it's important that each team knows what is expected of them. Information about the entire move should be provided to everyone involved. If the packers and shippers know what the receiving end of the move expects, they can pack items in the order they will be needed, minimizing the time it takes to get this back in place on the receiving end. Migration planners can coordinate the entire process by providing movers with a blow-by-blow runbook that details how items are to be palletized and loaded on the truck.

Engineered Live Migration

A "live migration" is achieved by mirroring the original site at a new location. This usually means installing all new equipment at the new location, mirroring services, migrating data and cutting over when the two sites are synchronized. A live migration, if properly planned and implemented, usually results in less down time than a forklift migration. Planning a live migration usually takes a lot longer, since testing and dry-run migrations are part of the process.

Partial Upgrade

Sort of a hybrid between a forklift migration and a live migration, a partial upgrades involves physically moving equipment to the new location and marrying it to new equipment at the destination. This can be more cost effective than a live migration but usually carries the most risk, since it may be very difficult to plan and test how the old gear will interact with the new equipment. There are generally more unknowns with partial upgrades. We have found that partial upgrades usually take longer to plan and implement since they are typically performed in several steps over a few weeks or months.

Vendor Coordination

Know Your Terms

Before moving anything, you should how vendor warranties are handled. Many SAN providers insist on having certified help on hand whenever their gear is physically moved. Check your support contracts and warranty information since many vendors will provide affordable help during physical migrations.

Who are Your Contacts

Your moving-day runbook should contain a detailed list of contact information. Gather names, contacts and as many phone numbers as possible for each vendor who will be involved with your move. You should not have to search for a phone number if something doesn't go as planned on moving day. Since you'll be dealing with down time and expensive hourly contractors, your contact list must be exhaustive, including even the local pizza delivery shop.

Scheduling is Key

Moving in most cases does not happen in one step. Each portion of the move should be planned on a time line to help minimize expenses that will be imposed by hourly employees on overtime and third-party contractors that will be working on an hourly basis. By knowing ahead of time when you need bodies improves efficiency on many levels. Too much help at the wrong time can be as troublesome as not having enough help when you need it.

Runbook Development

Document the Entire Move Before You Begin

Step-by-step documentation is the best way to ensure a smooth migration from one site to another.  We've briefly covered many of the things you should consider while planning your move. All the research and planning in the world will not help much if you don't have a fully-documented game plan as you move to the implementation phase of your project.

Build a timeline of events that will occur in the days and weeks leading up to your move. This is basic project management and will not only help to keep the project on track, but will provide a great deal of insight related to potential problems as you reach moving day. Your runbook will be especially important on moving day, as it will provide the hour-by-hour breakdown of what is happening, and how to roll back if you should suddenly become aware that you cannot proceed.

Delegate Tasks and Responsibilities

If you're lucky enough to have a skilled and knowledgeable staff at your disposal, by all means engage them in the planning stage. Teamwork is important and by asking for and valuing your team's input, you will put them in an ownership position. By getting your entire team involved, you may discover and eliminate minor issues before they become show stoppers. Assign individuals or teams that can assist with network planning, staffing, telecom circuit move coordination, and logistics.

Consider the three major functions of a move: Logistics, telecommunications and coordination. Logistics, for the most part, can easily be hired out to a contract mover. Major cities often have moving companies who specialize in, or at least have experience moving computers and related equipment. If you have a lot of telecommunications circuits, swinging those WAN and MAN connections can be a daunting and time-consuming task. Your full inventory and circuit audit, discussed early in this document, should have you fully prepared to place change orders with your telecommunications providers.

 There may be skilled help hiding in your organization that you didn't know existed. You can even work with other departments or coordinate through your human resources department to find helpers who can assist with racking and stacking equipment, driving trucks and pulling cable.

Have a Backup Plan

Finally, have a backup plan in case you need to delay the move. There are countless reasons why a move could be cancelled at the last minute. It could be as simple as a mandate from upper management, or could be a seemingly unimportant as losing one of your key employees due to illness. At any rate, you should have the option to postpone a move with little notice to a secondary date a week or two later.

Eric Hendrickson is President and CTO of Iophase Inc., IT network management outsourcers known for email spam and virus protection, email and web services hosting and datacenter management for small, medium and large business customers. Based in Irvine, California, Iophase uses and supports open source software and databases and is dedicated to the open source ideal. Mr. Hendrickson holds industry certifications in Linux and Cisco.

 

Forward this page to a friend

Your Email: *

Your Name: *

Send To: *

Enter multiple addresses on separate lines or separate them with commas.

Message Subject: (Your Name) has forwarded a page to you from NaSPA.com

Message Body: (Your Name) thought you would like to see this page from the NaSPA web site.

Your Personal Message: *

Your vote:

1

2

3

4

5

No votes yet