For Seattle’s cloud infrastructure community, September 4th, 2014 was marked by a peer-led DatacenterDynamics Converged conference in Bellevue, WA. With Schneider Electric as Platinum sponsor, the expo was filled with a related group of datacenter infrastructure providers and suppliers. The notes below are from Liana Antanovich, Nuvalo researcher and analyst.

The DCD Converged Conference and Expo brought together the people, the processes and the technologies in order to help our delegates develop a world-class data center strategy. Out of their multi-track conference program – headlined by a host of international experts – we chose the Design + Strategy track, which looked at “what’s new” in our industry, and debated “what should change.” By our estimates, the most educational keynotes were from Bill Hodges, Program Director, Providence Health & Services Datacenters, and from Greg Couch and Rich Stewart, IT Infrastructure leaders at University of Washington.

Greg and Rich gave an overview of their applied philosophy that enables them to successfully manage 2 owned and 3 leased datacenters, processing 12 Petabytes of UW data exchanged over the Internet. That is 31.7 million server hours a year for research and calculations that are being backed up by $100 million annual IT budget. The diversity of IT systems deployed is astonishing. According to Rich, UW still runs its payroll system in Cobol, and this payroll system has been running for the last 40 years. Greg calls out their challenge as “running Tier 2 infrastructure at Tier 4 uptime,” and outlined the core principles of facing this challenge, that at times contradict the expected norms of modern IT:

  1. Automate anything you can. Trust and enable automation. Identify, simplify and automate process and routine.
  2. Prioritize. Eliminate nuisance alarms/alerts to avoid customer service impacts.
  3. Drive bottom-up operational improvements.
  4. Don’t create policies and procedures that get in a way of performance or a good customer service.
  5. Minimize complexity.
  6. Hire good quality people.

Greg argued that procedures can lock you into failure. He advises to avoid creating environments for people with clipboards checking off items. Rather hire smart people, spend budget on people, not procedures.

The other piece of UW IT team success at Sustaining Datacenter Business Model is a P&L approach to management. “Manage datacenters as business,” says Greg. “Split staff and operations into two teams. First is facility operations, treat it as a cost center and administrative overhead – allocate budget, manage well and forget about it. Second is datacenter infrastructure as revenue center, with totally different team of engineers.”

Bill Hodges took a view of modular design in a data center. Providence rents space at Sabey Corp datacenters in Tukwila, WA for the last 5 years, managing their infrastructure to 99.99% availability. According to Hodges the future datacenter should be considered as a triad of facility, white space and technology. White space stands for the number of new server installs. Each element of the triad is to be looked at, designed, and managed as its own module.

What are the principles of a privately enterprise-owned or commercial datacenter strategy planning? Bill calls out a few that are cornerstones in his eyes:

  1. Business decision – what is our business? Is it collocation, vendor hosting, or cloud based offering?
  2. Efficiency – do what is most cost effective. Period.
  3. Manage datacenter as a business even when enterprise-owned. Treat it as internal collocation. Hence all the customer service and account management philosophy.
  4. View datacenters and technology as one. Few do that. Most prefer to conveniently limit themselves.
  5. View capacity as limited integrals, universal pools with universal access. Note: future capacity pools are modular, just-in-time, cross-functional. Maximize capacity pools to full potential.
  6. Integrate vendor relationships into workforce management.
  7.  Converge technology with integrated, best of breed automated solutions. In order to reduce capacity pool size: ports, cabling, hardware, licenses.
  8. Because core infrastructure life-cycle is minimum 10 years, you can really benefit if you:
    • simplify, reduce vendor diversity
    • standardize – be consistent like McDonalds
    • create quality process foundation, repeatable processes
    • aim for zero data center physical expansion – draw a line on the sand
    • deploy just-in-time capacity – hardware, software, people, customer
    • pursue investment multiplier – leverage invested dollar more than once
    • minimize cost

From the idea of ‘disposable data centers’ to ‘software defined ‘ and everything in between, DCD Converged is sure to keep North America’s most senior data center professionals up to speed with an industry changing at lightning speed. According to recent conversations, in 2015 this conference will be held not in Seattle but in San Francisco, CA.