In March 2006 the Census Bureau awarded a $600 million contract to a systems integrator to provide the IT infrastructure, support services, hardware and software to support the 2010 census field data collection.
We hear that the Census is preparing to award a contract to a value-added reseller, to build and maintain infrastructure for the 2020 census field data collection. But this time, all that infrastructure will be in a public sector cloud. It looks like we are now beyond the hype curve and into the reality of cloud service providers delivering enterprise class infrastructure as a service. In this case, the reseller will provision and maintain these services on the IBM SoftLayer FedRAMP cloud. We don’t know the contract value yet, but we bet it is substantially lower than in 2006.
Consider the shifts apparent in these two awards:
- From on-premises government iron to the FedRAMP cloud.
- From a systems integrator, Harris, to a cloud service reseller and implementer.
Census avoids the costs of building its own infrastructure by purchasing a utility service. And most of the systems engineering required to produce the infrastructure was already expended by the cloud service provider, leaving lighter provisioning work that can be performed by a reseller.
The choice of cloud service provider is also significant. SoftLayer mitigates the compromises that typically go with a shared cloud. They provide a single tenant environment operating in a US government-only cloud. Customers have the ability to build physical or virtual machines up from bare metal with any software, so they can “lift and shift” their applications as-is, avoiding the cost of converting or rebuilding applications to work in a proprietary cloud. And they only have to pay for what they actually use –by the hour or by the month.
The lesson is that Census Bureau specified the outcome they sought –scales, service levels, and security — not the labor hours or the type of infrastructure they wanted to see. By focusing on the outcome, rather than the engineering, Census will get more service for a lower cost. If an agency solicitation lists the labor categories they want to see, and the level of effort they expect for each category; and/or if it specifies the engineered infrastructure they want to use; then the agency can expect to get more of the same – costs, levels of service, levels of security, and levels of complexity just like they have those today. But if an agency wants to get a lot more for their money, it should write the solicitation to specify the outcomes it wants – the service levels on their key performance metrics — and let industry figure out how to deliver those for the lowest price. The outcome may be surprising: a wholly different kind of winner with a wholly different solution – at a vastly different price.