Notes from the Consultant’s Jungle

The past two years have brought an explosion of activity in the area of data center facilities.  On the provider side, we can talk about services we expect from the Cloud.  The proliferation of XaaS, exponential growth of content for Web 2.0 services, and the interest in Cloud-based services by the Enterprise drives demand for storage and utility-style computing services.  Providers of these services are expanding their facilities’ footprints at an impressive rate.

For the enterprise, Business demand for IT resources has stressed the resources of enterprise IT facilities both above and below the white space.

At the turn of the century, it was common to plan IT facilities using the Watts per Square Foot (W/SF) model.  We saw data centers operating comfortably at 50 W/SF in those days, and if you had a generator backup you felt pretty good about availability.  Things have changed drastically.

We now talk about the tier-level rating of data centers to align facility availability with the operating model of the Business.  We focus much attention on the TIA-942 and Uptime Institute guidelines for redundancy and topology of MEP systems.  Instead of the W/SF planning metric we use kilowatt-per-rack (kW/rack) because it’s much more relevant to how we deploy technology today.  The density of silicon in the data processing footprint is constantly increasing, which is driving power and cooling demand of the facility.

In 2003, we saw power density levels already at 4kW/rack.  By 2005, when blade servers first began to be a common site in the data center, 9kW/rack was common.  Today as those systems are seen as more of a standard building block, 20-50kW/rack is becoming common.  An enterprise IT facility built before 2005 very likely cannot support this level of demand.

Still, the growth of the Business and evolution of data processing systems is outpacing the capacity of IT facilities, and something must be done in order to continue to scale the IT capabilities.

Which way to turn?

Tough choices must be made by the technology executive to support Business growth and good corporate due care.  The universe of choices typically is:

1. Expand in place- This option typically involves adding MEP infrastructure to support higher power and cooling demand and include redundancy to support the necessary availability. If floor space capacity was not already a problem, it usually becomes one as the additional MEP capacity also has floor space requirements. Depending on how the cooling is delivered, it’s possible that the building architecture will not allow for the required air flow (e.g., sufficient slab-to-slab clearance).

This option can be complicated further if the existing facility is a leased commercial building.  The landlord or the covenants of the office park may not allow for the changes necessary to accommodate the needed improvements.

2. Lease space or acquire another data center- Find a data center that fits your needs but has been abandoned by someone else, or find core/shell space that can be outfitted as critical IT space. This option is often investigated as a primary directive once the option of expanding in place is ruled out, or because of the desire for a higher level of good governance and risk mitigation as opposed to Option #1. This option is very difficult to pull off, in spite of sounding like a really great idea.

First of all, most of the vacant data centers you will find (if you’re able to locate any in your area at all) have likely been built five to ten years ago.  As such, they are likely to be under the same MEP capacity constraints that caused you to look elsewhere in the first place.  That is to say, vacant IT facilities on the market are probably built for 30-50 W/SF power densities and using Tier 1 or Tier 2 MEP topologies.  This probably doesn’t help your cause.  Additionally, the slab-to-slab distances in such facilities are usually preclusive to increasing the MEP capacity to meet most contemporary enterprise needs.

Secondly, when looking for a core/shell building to renovate into a data center, there are many devils in the details that complicate finding a good candidate property.  There is a tricky formula of interdependent variables to finding a building that will make a good data center.  Among these are:

• Location- A location that has the setting and surrounding space to mitigate risks associated with traffic, intrusion, and so on.
• Access to Fiber- Cost effective access to telecom facilities
• Slab-to-Slab Clearance- Sufficient floor to ceiling distance to allow installation of the raised floor, cabinets, overhead cabling, fire suppression, and lighting. This factor has killed many deals in my own experience.
• Column layouts- the frequency of columns in the room space will impact the efficiency yield on the IT floor.
• Zoning- Permission to install outdoor generators, chillers, and storage of diesel fuel
• Building structure- Architectural characteristics to minimize risk to the facility such as window location and wind rating of the roof
• Soil- Water tables, flood zone, waste water, fill dirt, bedrock, and so on are issues that need to be resolved through soil testing.
• Lease vs. Buy- The tradeoff between buying and leasing, even with the significant costs of improvements, can be a tricky one.

Be mindful also that even if you stumble across the perfect building for the data center, it’s the MEP infrastructure that drives the cost.  Finding the right building saves only a small percentage of the overall development cost.  Most often, one will find it’s easier to buy land and build a completely new facility than to fit into an existing structure.

3. Outsource to 3^rd party provider- Move your IT infrastructure and operations to a managed services provider… also known as “your mess for less.” This option is a good fit for enterprises that decide that support of IT is not a part of the core business AND for which there are already mature technical operations processes. Take note of that second point. If the technical operations, configuration management, change control, and general technology governance are not in a mature state, the outsourcing of that environment will not be a smooth endeavor. It’s true that outsourcing to a company for which this is a core competency (e.g., IBM, EDS/HP) is a way to achieve an improvement in technical operations, but it’s rarely a smooth transition unless your own operations are at least in a decent state of maturity before the move.

Outsourcing of large enterprise data processing environments is costly to plan as well.  One should plan on a significant investment of staff resources and/or consulting professional services to prepare for such a move and to vet candidate providers.  In either case, it is very important to have a crystal clear picture of the TCO over the life of the agreement because firms are often surprised to find that outsourcing does not necessarily reduce expenses and that once the environment is moved it is very difficult to pull back in.

4. Collocation or Managed Hosting- Move your data processing equipment to someone else’s data center, and manage it remotely. This is a great option to singularly solve the problem of exhaustion of facility capacity. A collocation provider can be selected that has a facility meeting the desired tier level, and as a tenant in the facility one can amortize the costs of that hefty MEP infrastructure across the fellow tenants.

If you’ve not been shopping for collocation space recently, you’ll probably be surprised by the change (i.e., increase) in pricing.  Collocation providers once used the rack or cabinet as the base unit of cost to the Customer.  Because of the rise in utility rates and the exponentially increasing demand on power and cooling by contemporary data processing equipment, providers have increasingly moved to a metered power unit of determining the cost to the Customer.  In other words, your costs will be determined primarily by the amount of power used.

Your options are better too if you’re not too concerned about where your equipment will be located.  Remote Hands services can usually be included in the hosting agreement and enables a broader selection of providers and prices when geography is not a driving issue.

The availability of specific telecom carriers and network peering is an important consideration in the selection of a provider and facility as well.

Furthermore, if your data processing infrastructure is highly virtualizable on commodity equipment, then a Hosting provider can also be included in the candidate list.  With a Hosting provider, one moves the application and data to the provider, where it is hosted on shared equipment that is owned by the provider and not to the Customer.  For some, this is a great option that eliminates the need to plan capital expenditures on equipment refresh.  For some, the application architecture obviates this option as a possibility.  Others still are sometimes simply morally opposed to running their data on anyone else’s equipment.

5. Build new- Build a new data center, positioned to accommodate your unique projected growth plans and create a high-value real estate asset for your company. Yes indeed, every IT Director’s dream is to create a shiny new high value asset to add to the company’s portfolio and start fresh with capacity to spare. In this second half of the first decade of the 21^st century, it’s not uncommon to see this done. Indeed we are in a data center construction boom.

We have written in other posts about the costs of building data centers, and the fact that the MEP infrastructure is by far the dominant cost driver.  The land and the building costs will pale in comparison.  Besides those costs, one should be aware of the lead time involved with that MEP infrastructure.  At this writing, lead times of 50 to 70 weeks are common for large generators (the big boys like Google and Microsoft are buying them by the dozens, and the reconstruction efforts in Iraq are consuming much of the production of generators as well).

These projects are not only expensive but very complex, requiring specialty consultants for planning and selection of contractors, real estate, zoning, tax incentives, the complexities of mission critical facility construction projects, and the logistics of commissioning and relocation into the new facility (yes, after all this, there’s still a data center relocation project to conduct).  Further details on this will be covered in other posts and in the book.