A data center checklist for facility design and IT ops

Computing equipment

Many organizations choose to work with data center providers that supply computing equipment. Others handle equipment selection, install and maintenance in leased or owned space.

• Computing equipment compatible with existing applications

Survey the current suite of servers, storage arrays, network switches and other IT equipment. Calculate the available computing resources and verify the gear is fully compatible with, and has the capacity to support, your existing applications.

Legacy or custom applications with hardware-specific requirements may cause problems on hosted servers. Unique operating systems and hypervisors pose compatibility problems as well. See whether an outsourcing provider will supply alternate equipment or support owned equipment from your IT organization.

• Fast equipment refresh schedule and detailed roadmap

How old is the current data center compute infrastructure? When was it last replaced? When is it due for replacement? What new equipment is under consideration? These insights will guide server consolidation and workload balancing plans.

A data center provider should offer transparency into its technology refresh plan since your business will be running on its equipment.

• Coordinated equipment refresh process

Equipment refresh processes are disruptive, so a provider must communicate update and upgrade plans to users and work in conjunction with them to mitigate the consequences of any equipment work.

Will workloads keep running during equipment updates? How are unavoidable disruptions communicated? Do you have a plan for these disruptions, such as preemptive backups or temporary workload relocation?

• Experienced local staff and expert support

How many IT professionals are on staff 24/7 (even if the staff is simply a worker who follows instructions from your team remotely)? If the data center is understaffed or technicians are only on call, disruptions could hurt critical applications.

What is the service response time? Are escalation paths clearly defined in service-level agreements (SLAs)? Use general calls to the support team to test response times and service quality.

• Management and performance tools

How is system performance measured and monitored? If you're dealing with an outsourcing provider, are you privy to all or some of this information?

Typically a hosting company won't disclose overall environment data, but if you're renting equipment, consider a Web portal or application performance monitoring and management (APM) tool. You may need to set up your own APM to ensure SLA compliance.

• High network bandwidth

The network bandwidth should be verified and spelled out in the SLA. For data center providers, 10 Gbps is common. Include network bandwidth in ongoing performance monitoring.

Can you burst to more bandwidth, and how much will that cost? On the flip side, can you throttle bandwidth to control costs, or balance workloads across multiple sites to prevent overusing one?

Security and access control

Physical tampering and thefts leak data and expose the company. Security may fall to the data center operator, IT organization or both.

• Physical security

Is computing equipment secured in the facility? Options include unidirectional mantraps at the data center entrance, locked cages around private equipment areas -- each with unique physical or electronic keys, and locked access to areas where private data travels, such as cable troughs and telco interface areas.

• Access documentation

Can you track employees, contractors, suppliers and visitors? Note the method: sign-in sheets, comprehensive camera recording, electronic access badges or biometric screening.

Is there a permanent record of anyone who could touch your equipment, cabling troughs, patch panels, or any area associated with your workloads and data? This is especially important with multi-tenant equipment, when a technician's work on one server might affect multiple clients.

• On-premises access staff

Some data center providers go beyond electronic surveillance and include full-time staff -- and even employ armed guards -- to enforce visitor and vendor tracking, facilitate 24/7 data center user access, and interface with law enforcement and other security protocols.

• Secure equipment and/or data destruction

Retired or repurposed storage components easily leak data. How are old, obsolete or failed disks secured or destroyed? Are disks tracked when they're removed from local servers or storage arrays and stored securely? Who destroys the disks, and if an outsourcing partner handles it, do they provide written certification?

Facility considerations

The right building can make all the difference when it comes to reliable long-term service, even if you're not actively in charge of facility management.

• Suitable safe location

Selecting a data center is all about location. The building should be clear of common environmental hazards like flood plains and earthquake zones. The site should also avoid proximity to manmade hazards like industrial centers -- oil refineries or chemical factories, the path of major airport traffic patterns, and major highway exchanges or railroads.

• Convenient shipping/install

Evaluate any features that expedite equipment deployment and replacement. For example, secure and sheltered shipping/receiving docks let workers load and unload IT equipment without it getting wet or snow-covered.

Multistory buildings need freight elevators close to the docks. Secure freight access to the data center should be large enough to accommodate oversized equipment, such as full-height racks with in-rack cooling units. It's also helpful to have access to a staging or "burn-in" area where equipment can be preassembled and tested before moving into production.

• Office amenities

If you're supporting staff on-site, be sure that the facility has some dedicated office, canteen or other human-friendly space to work in, away from the heat and noise of server fans. Niceties like showers or nearby hotels benefit employees in a remote data center facility.

• Environmental control and safety features

Survey the building for state-of-the-art smoke detection and fire suppression systems that rely on dry chemical extinguishers within the data center. The facility should use modern electrical grounding standards, such as PANI grounding, to ensure safety.

Since many facilities use water for cooling (such as water-based chillers or heat exchangers), comprehensive leak detection is critical.

• Ample power and resiliency

Power is an increasingly scarce and expensive commodity that is not equally available from one place to another. Select a data center in a region where energy is relatively inexpensive and plentiful; regional energy shortages can translate to premium power costs.

Look for secondary utility providers feeding the facility from independent substations, backup power capabilities such as generators, or proximity to alternative power sources like wind farms. Industrial-grade, always-on fuel cell generators, like Bloom Energy box arrays, may require close proximity to natural gas or other biogas sources.

• Cooling and resiliency

The data center probably uses a mix of conventional mechanical air conditioning and chilled-water heat exchangers, but even a short cooling outage can impair operations. Verify that redundant cooling units are available and access redundant power sources.

Cooling capacity should also allow for future growth. Humidity control systems must maintain a comfortable relative humidity level for human occupancy and electronic safety.

• Standards compliance

Verify that you'll meet any specific standards or requirements imposed on data center facilities for your industry, with documentation. Standards that involve data centers include SAS 70 Type II and PCI DSS.

• Network connectivity and carriers

Find out which carriers have connectivity to the data center, such as Cox, AT&T, Verizon and others that may vary by region. A neutral colocation or hosting provider should support a variety of regional carriers and provision for redundant, independent connectivity through several different fiber entry vaults.