1. Gather requirements from documentation
2. Create specifications for each requirements
3. Decide to issue RFIs (or not)

GOAL

Data Center Migration - build new data center - Move existing services

IT Details

Dimensions

The new machine room measures 36 feet (11 meters) by 23 feet (7 meters) and is 10 feet (3 meters) tall. Those are raw measurements without a floor or ceiling.

Flooring

Operations prefers a raised floor be used for the distribution of HVAC and electricity to racks. The raised floor in the old machine room was 12 inches (30 centimeters) tall and was generally considered sufficient. None of the current staff remembers any issues with the raised floor other than there was never a Floor Panel Lifter around when you needed one. One operations manager indicated it was important to use floor panels that matched the width of the equipment racks.

Ceiling

Operations desires that cables for data distribution run through cable trays that hang from the ceiling. Several people interviewed warned about using light or inexpensive trays as there was an incident years ago when a tray became overloaded and failed. One person mentioned there were manufacturing standards for cable trays and ladders. They also suggested that loads might be lighter in future years as technologies in the machine room continue to evolve but that you can never rule out the possibility somebody might carelessly hang something heavy from a cable tray.

Operations did not express requirements for the ceiling material other than a preference not to use an open ceiling. They suggest some form of hanging ceiling allows for better control of HVAC, but they realize fire suppression systems might also have requirements that are more important than their preferences.

Electricity

Operations desire high availability from the electrical system. Grid power, battery backup, and generator power have all been used in the past to collectively provide 99.995% availability of power. The current machine room was designed to provide 2 hours of battery life without being fed from the grid or the generator. As the number and size of computers has decreased, that battery life would have been greater, but was never measured. FiberTech has a sustainability plan that calls for increasing use of renewable sources, but for the purposes of this project that should be considered grid power. Management is not yet ready to abandon a fuel-driven generator approach, specially since it is rarely used and already owned. The current generator is 10 years into its expected 20 year life. It is unclear if it makes more sense to install a new generator or move the existing one, or even if the existing one has to move. You sense that management would like not to move it, but that running new circuits will ultimately involve some down-time to essential systems. Installation of a new generator would allow for dual operation, if needed. Senior management recently expressed Business Continuity worries related to hurricanes. This suggests the company has little risk tolerance for extended outages.

Heating and Air Conditioning (HVAC)

Operations desires to keep the data center between 66F and 76F (19C and 24C). Actually, a range down to 64F (18C) and up to 80F (27C) is acceptable, but temperatures in that extended range are considered a warning level and require some kind of response. Operations desires to have at least one spare HVAC available beyond the required number to allow for a device failure or to enable faster temperature/humidity recovery, if necessary. Nobody in Operations remembers ever having to provide heat to the data center, but suggest that it might be necessary. Humidity is to be kept within the range of 45% to 55%, but 40% to 60% is actually allowed. Again, the inner range is desired and the outer range sets warning levels.

Fire Suppression and Related Controls

There is probably a standard/requirement for emergency lighting to be met. Operations does not know what that standard is. They prefer to use some kind of gas suppression system so that they could reuse the computing equipment after the fire, but they recognize if a gas system fails to stop the fire, then the rest of the building gets put at risk so they would be okay with a water-based system if that is what the company decides. They point out that if a water-based system is used it becomes important to depower the facility.

Security

There must be two entrances to the data center. One will be a normal entrance accessible via a key or a key card. That entrance will lead to a small unmanned control center, which is expected to act as a sign-in area, temporary office, and tool storage.

The other entrance will have access to an existing elevator that connects with a loading dock. This loading dock entrance should only be able to be opened from the inside.

Both entrances should be functional emergency exits. Cameras must provide surveillance of both entrances and every corridor between racks within the machine room. Operations desires that footage from the cameras be available for 7 days, but would settle for 4 days as that would allow for review after a long weekend or over most holidays. They do not care if the cameras are motion activated or not.

Operations desires to save energy by turning off most lights automatically when nobody is in the data center. Either the lights should be motion activated or the lowest lighting should be dim yet bright enough that a person should be able to safely navigate the data center in order to turn on lights. (This would be a precaution in case somebody was caught in the data center when somebody else turned out the lights.)

Rack Installation

• 6 feet (180 centimeters) tall by 2 feet (61 centimeters) wide by 3 feet (91.5 centimeters) deep.

Networking for New Machine Room

The Internet Point of Presence (POP) currently resides in the old data center. New cabling will have to be run and a plan for migration will have to be made. The Networking team does not plan to reuse the existing routers, so new firewalling, routing and switching can be installed in advance of a cutover.

Connections to the Warehouse, Factory, and the Boutique stores are all based on MPLS connections. Which both the Networking team and the CFO want to revisit.

The Information Security team does plan to move its newly acquired turnkey Data Loss Prevention (DLP) system to the data center. It involves two redundant systems, which can be moved from the old data center to the new one separately such that one is functioning at any given time.

The Information Security team also has a number of Snort boxes, which it uses to monitor various network segments. They are able to build and deploy new boxes to the new data center's internal network without having to tear down support for the old data center's internal network.

The new data center's internal network will largely support a virtual machine cluster. Within that cluster, there will be a number of service clusters that everyone wants to keep isolated from each other for both security and practical reasons.

The networking team keeps promising to deliver to you a network diagram for the current data center, but the results keep getting delayed. While it would be handy, the new data center isn't required to reconstruct the past pattern.

Service: DNS

The data center houses an internal primary domain name server, two internal secondary domain name servers, and two public secondary domain name servers. These DNS servers are located in the virtual cluster. The primary DNS is only visible within the data center network, with access restricted via firewall to only the secondary DNS and is not accessible to the rest of the company nor any external requests.

Service: Corporate EMail & Calendar

Tech outsources corporate EMail and Calendaring. Those services need to be appropriately exposed through the border firewall to the vendor, Gigglesoft.

Service: External Cloud Service Providers

Thus far, Tech has only undertaken a number of temporary pet projects involving Infrastructure as a Service (IaaS) and Platform as a Service (PaaS). These efforts have largely concluded that in general it is still less expensive for Tech to run On-Premises services, but that a hybrid approach would provide some valuable flexibility for which the company is willing to pay.

Service : Customer Relationship Management (CRM)

Tech operates a Commercial Off The Shelf (COTS) Customer Relationship Management (CRM) system that has been hosted on physical machines. The team that supports this service desires to migrate to virtual machines both for more operational flexibility and in case it would eventually make sense to move to cloud hosting. The hosts involved are:

• Web-based Customer Service Frontend (Public)
  • 2 load balanced web servers
  • Only 1 web server is necessary to satisfy peak demand.

Service: Financial System

And, perhaps the most significant change taking place is that there is another project underway to plan for a migration to a new Financial Management System. The current system was purchased years ago by a small vendor who has announced they will be going out of business in two year's time. The owner is ready to retire and no other company is interested in buying the company. Some of his employees have expressed interest in maintaining support for the product, but the few remaining customers have largely seen the writing on the wall -- it's time to manage a nice transition instead of procrastinating and rushing through a rough transition.

The preliminary version of the plan for the new system is to acquire a more-capable replacement system off-the-shelf, get it functional as a replacement, then switch. If the new data center is far enough along the new system could be assembled there, otherwise, it is hoped that development and testing in the existing virtual machine environment would make deployment to the new data center relatively painless.

While a new financial system means lots of potential changes to financial operations, any additional effort for IT seems to be largely in figuring out how to extract and transform data from the old system and load it into the new system. (ETL = Extract, Transform, and Load.) In particular:

• Inventory (old COTS) - ETL
  • Headquarters (Raleigh, NC)
  • Central Warehouse & Manufacturing (Charlotte, NC)
  • Boutique Stores (Cities in USA and Canada)

Management is certain the existing support team for the Financial system does not have the cycles (time) to maintain support for the old system, learn the new system, support the data center migration, and perform the ETL work. Management has not yet decided if the ETL effort should be outsourced or if contractors should be retained. At this time, it looks as if senior management is leaning towards a mix. They would use contractors to "backfill" the existing team, freeing up some of them to work on this project but still outsource the ETL part of the work since that effort would have clear objectives and there would be no need to acquire and learn an ETL tool.

The Future

The data center has long been the heart of the Chief Information Officer's (CIO) operations because that is where data services for internal customers have been located. While moving services to the cloud has been in vogue and certainly has gotten a lot of press, the CIO knows associated costs do not currently support such a move, except in a few cases. He would like to be able to transition his services to the cloud in times of non-routine need, such as to handle a prolonged surge in demand or to provide business continuity in the face of an unanticipated event.

The CIO is also aware the data center, and the Operations team, might eventually become a resource shared with the Chief Technology Officer (CTO) who currently leads the Research and Development teams as well as the Marketing team. Other than the corporate website, Internet-based marketing has normally been handled through external providers.

The CTO is not necessarily focused on the data center at this time, but likes to keep options open. The CTO suggested appearances might be awkward if they were selling t-shirts that were IPv6 capable, but the corporate data center was not. The CTO also suggested the integration of computing into clothing could lead the company into handling medical data. There might soon be a need for something like a personal portable network of clothing sensors back to a HIPAA compliant data center over a VPN tunnel through a cellular phone router/firewall. They made it clear they were not suggesting the data center needed to be HIPAA compliant, but that it would be best if the data center design did not preclude that possibility.