HyperSpeed: How the Data Center Supply Chain Keeps Pace

ERIC BOONSTRA, Iron Mountain

Eric Boonstra: It’s mainly a race against the clock. Who has the capacity to deliver the huge demand and huge volumes that hyperscalers are asking for (10-20 MW) at the right time in the right place? And do you have the right infrastructure and the flexibility to serve hyperscale and retail customers in your campuses.

In order to win the race, you need standardization of build and processes. You need to have your supply chain standardized and ready to go and contracts with your suppliers in place. You need to be able to respond immediately on customer needs and changing requirements and have flexible infrastructure. The companies that have all of that will be the leaders.

ERICH SANCHACK, Digital Realty

Erich Sanchack: We see two areas we need to address to keep up with the growth in the market. The first is the definition of high-density. With emerging applications like AI and machine learning becoming a mainstay in today’s enterprise, we have to rethink what makes the data center AI ready and if it can support high-density workloads. From the power management, to new cooling requirements to keep the data center stable and running, many facilities today are not up to the task. This challenge is leading many to look to their data center partners for a purpose-built infrastructure ready for advanced computing applications.

Another factor to consider is that many of the digital-native businesses are looking to minimize their environmental footprint and want access to sustainable energy in the facilities they select. Digital Realty has already taken a number of key steps on this path and is already fueling its US Retail Colocation business on renewable power.

Finally, we can’t ignore the people aspect of our industry. Today the talent required to build advanced facilities is incredibly scarce. The staffing needs of data centers are very complex, ranging from facility managers to manage the power supply and cooling, to those that manage the networking technology to enable better data flow. With many that started at the advent of the industry retiring, providers need to invest in training to skill up new workers coming in, and automation to automate routine tasks and augment the abilities of the existing workforce to keep up with demand.

Mitch Fonseca, VP, Data Center Products, Cyxtera

Mitch Fonseca: As discussed in our Executive Roundtable response regarding international demand , to speed data center deployment in new markets Cyxtera is expanding our footprint by leasing existing facilities while planning new builds in strategic locations. Cyxtera partners with hyperscale providers with state-of-the-art data center designs and operational excellence as well as a common interest in offering customers flexibility, security and connectivity. This results in a better ecosystem in which retail colocation customers and service providers can connect directly to hyperscalers within the same facility.

Whether the facilities are built or leased, one of the most important issues impeding the pace of deployments is the 3 to 6 months it typically takes enterprises to provision colocation environments within that facility. Our customers – including enterprises, service providers and technology companies – have compressed that timeline down to hours and days, not weeks and months, with Cyxtera’s CXD on-demand provisioning of interconnection and preconfigured dedicated infrastructure. CXD’s ability to seamlessly expand and integrate non-contiguous capacity by stretching layer 2 networks within a facility or across a metro region also maximizes utilization and significantly accelerates deployments.

Jeff Klaus, General Manager of Intel Data Center Management Solutions

Jeff Klaus: Software defined capabilities become really attractive when you’re looking to efficiently utilize constrained resources. Also from within the server there are some great examples of how to optimize:

• Looking at newer server OOB technology to drive better datacenter utilization of power and space from real-time telemetry.
• Developing optimized memory strategies (for DCPMM) in program design will benefit cloud and HPC performance.
• Implementing compute utilization strategies (SST) to right size need dynamically.

NEXT: What lies ahead for data center automation?  

Keep pace with the fact-moving world of data centers and cloud computing by following us on Twitter and Facebook, connecting with me on LinkedIn, and signing up for our weekly newspaper using the form below: