Census Bureau Counting Heads in the Cloud
The U.S. Census Bureau is singing the praises of cloud computing.
Census is taking advantage of several cloud-based computing services -- from content delivery networks to hosted applications to free Web-based services -- for its decennial survey.
Census CIO Brian McGrath says the bureau has had a great experience buying software and infrastructure as a service, and that this approach has been an efficient and cost-effective way to meet the peak processing demands from the 2010 Census.
"We use the cloud in eight specific instances around the decennial survey," McGrath says. "That provided a huge benefit for us because we didn't have to stand up an infrastructure. We knew our requirements were for a definite period of time."
The Census Bureau's positive experience with cloud computing comes at a time when U.S. government agencies are being encouraged by Federal CIO Vivek Kundra to embrace cloud computing as a way of saving taxpayer dollars. Supporting Kundra's position, a recent Brookings Institute survey estimated that government agencies can save between 25% and 50% by using cloud-based computing services instead of internal IT resources.
Industry observers say many agencies like Census are interested in building their own private clouds.
"Fear of information being made available over the public Internet is keeping federal agencies from wanting to use the public Internet as the cloud," says Susan Zeleniak, group president of Verizon Federal. "They're going to want to use private clouds. That's what we see more."
Census said it spent $11.8 million altogether on the eight cloud computing efforts that supported the 2010 Census.
In January, Census began using Akamai to enhance the performance of its redesigned Web site -- www.census2010 -- which features video clips, blogs and other interactive content aimed at citizens. The new Web site attracted 4 million to 5 million hits a week at its peak, about double the traffic of the bureau's legacy Web site -- www.census.gov -- aimed at statisticians.
"For our new Web site, we went to the cloud," McGrath says. "We went with an infrastructure-as-a-service solution, and what a great experience that was. We contracted with Akamai to use their CDN…At the peak of our usage, we were servicing somewhere around 85% of our content from the edge, and it was not even coming back to our infrastructure."
McGrath said using the Akamai network provided a better-quality Web experience to citizens for less money than building their own network. Akamai also provided a barrier against distributed denial-of-service (DDoS) attacks.
"That was a huge concern for us that in the height of the decennial activity if we were a target of a DDoS attack or the site would go down or the performance would go down that it would reflect negatively on the Census Bureau and deter citizens from participating," McGrath says. "Using the CDN was a huge positive lesson. I don't know if it could have gone any better."
Census also used several software-as-a-service (SaaS) providers, including RightNow, which provides self-service customer support such as searchable FAQs. Census says it was able to get RightNow up and running 25 days after purchasing the system. Census says it would have taken six months just to select the IT infrastructure required to run the application in house.
Census uses GovDelivery, which provides outsourced e-mail delivery services to public sector clients. GovDelivery's built-in blogging tool was used by the Census Bureau Director to publish a blog within days of buying the service.
The bureau's Integrated Partner Contact Database is built upon Salesforce.com's platform, which it paid for on a subscription basis. Census was able to tweak the configurations on Salesforce.com software, rather than having to conduct any custom programming.
Census also is using the free Google Map's API to quickly develop mapping applications including an assistance center lookup and an interactive road tour.
To speed up acquisition of these cloud-based services, Census partnered with other federal agencies including the National Institutes of Standards and Technology and chose SaaS vendors that had already been certified by another agency. NIST is leading a federal cloud computing advisory council that is setting cloud standards and certifying cloud-based service providers to make it easier for agencies to buy cloud computing services.
"We didn't have to re-certify and re-accredit the systems, and it really pushed the delivery of the service down from months to days or weeks," McGrath says.
McGrath says the bureau is looking to expand its use of commercial cloud-based computing services where appropriate, and is also leveraging its experience with these vendors to build the Census private cloud.
"We have a pretty aggressive internal cloud effort that we are building out," McGrath says. "There are still some concerns about the security in the public cloud. I have every confidence that those will work out in coming years. For us, [the plan] is to leverage the efficiencies of cloud technology and build an internal cloud."
One reason Census can move so aggressively into cloud computing is that it has been migrating to virtualization over the last 18 months. As of June, the agency had 427 virtual machines running on 57 server platforms. Census uses VMWare as its virtualization platform. The bureau said it has spent $6.1 million on the hardware and software for its Windows virtual farm.
"We've highly virtualized our Windows environment," McGrath explained. "We've gone from a model where we had one application on one server. Now we've got hundreds of guests in our virtual farms, and we are realizing significant savings of $2 million a year because we've compressed down our hardware footprint."
Next up for Census is virtualizing its Linux servers, which are standardized on RedHat. "We're doing a cost-benefit analysis," McGrath says. "It looks like of our 1,000 Linux servers, 80% are very good candidates for virtualization because they are probably running at 20% utilization or less"
Census also is looking at homogenizing and virtualizing its storage platforms, which contain more than 2.5 petabytes of data from the decennial census and other regular economic surveys that the bureau conducts.
"Virtualization is a piece of the overall cloud architecture," McGrath says. "It's a logical first step because what it allowed us to do is to really show in a limited investment, in a limited scale, the benefits of the cloud…We've been able to demonstrate to our customers that we are able to reduce our footprint, we're able to provision services more efficiently with less operations and maintenance costs, and our security costs are reduced because we can do security at the architecture level."
Read more about data center in Network World's Data Center section.