RightScale Case Studies

Center for Biomedical Informatics at Harvard Medical School Minimizes Cloud Computing Management Using RightScale

Situation Overview

Dr. Peter Tonellato established the Laboratory for Personalized Medicine (LPM) in 2008 at the Harvard Medical School (HMS) Center for Biomedical Informatics (CBMI). The LPM's goal is to explore novel methods to promote individualized health care, to use genetics and predictive and preventive medicine to maximize patient benefits, to reduce overall healthcare costs, and to minimize patient's risk to adverse health outcomes. This challenging agenda involves identifying and translating solutions to the pragmatic hurdles of integrating new molecular devices, genetic and proteomic data, environmental factors, and disease-focused knowledge to point-of-care practice. The success of this agenda depends on the ability to implement systems that improve patient care, ensure data integrity, can be easily updated, and are minimally disruptive to the current health care workflow.

In order to handle the diversity of projects and large amount of data required by the LPM, Dr. Tonellato and Dr. Vincent Fusaro, a Research Fellow in Dr. Tonellato's lab, used cloud computing to build an on-demand computational infrastructure. By tapping the enormous power of cloud computing – in this case, Amazon Web Services' (AWS) Elastic Compute Cloud (EC2) – the LPM benefited from a large, cost-effective computing resource to support its translational research.

Project Challenge

The LPM needed a way to increase the efficiency of its cloud deployments. While cloud computing offered a powerful computational platform for the LPM, a major challenge was to reduce the amount of time spent managing the computing technology in order to increase the time spent on research and discovery. A major goal of the LPM's cloud management team, led by Dr. Fusaro was to create a low cost cloud-based Computational Center, while minimizing resource management and administration time.

By automating our cloud computing deployments with RightScale, we accomplished our goal to significantly reduce overall administrative and management requirements of our biomedical research cloud computing resource – and increased the time we devote to translational science projects.

Dr. Peter Tonellato, Director of the Laboratory for Personalized Medicine (LPM), Center for Biomedical Informatics (CBMI), Harvard Medical School

In order to better understand and maximize the use of cloud computing, Dr. Tonellato and his team organized and ran a "Translational Science in the Cloud" seminar. Participants in this Harvard seminar, or "palaver," conducted a series of scientific projects designed to explore the utility of cloud computing for translational research. The six research teams that participated in the palaver were geographically dispersed across the United States and Japan, with participants from organizations such as Harvard Medical School, Children's Hospital of Boston, and the Tokyo Medical and Dental University.

RightScale Solution

The LPM cloud management team selected RightScale as their cloud management platform because it operates at a level above the underlying cloud infrastructure and provides more flexible configuration options for launching servers. This meant less time devoted to instance administration and configuration and more time on actual research projects.

With RightScale, users manage deployments, which are clusters or groups of servers that work together, rather than individual servers. Deployments bring all the servers associated with an application environment together under unified management, saving time and reducing errors.

The RightScale platform creates a more productive environment for managing cloud deployments, providing best practice deployment architectures with complete automation to dramatically reduce complexity and administrative burden.

"Without Amazon's AWS and RightScale it would have been impossible to work collectively in one environment with such a diverse group, in different physical locations, and accomplish what we did in three months," according to Dr. Tonellato. "I like to think of Amazon Web Services and RightScale not just as a collection of CPU cycles and storage, but as a flexible computing resource and environment to manage diverse collaborations."

Using the RightScale Cloud Management Platform, the Harvard palaver project collaborators were able to launch and manage their project resources under sub-accounts. This greatly reduced the time the LPM cloud management team had to spend launching and configuring the desired number of instances for each project. RightScale also helped centralize the management by providing an overview of all running instances. "RightScale was critical in managing this kind of work," relayed Dr. Fusaro, who used the RightScale sub-account feature to set up each palaver project group with its own account to manage project resources. "We estimate that we saved 80 percent of our management time by using sub-accounts."

"Sub-accounts were my favorite feature in terms of organizational capacity and potential in a multi-institutional setting. With RightScale, we saw how powerful organized and tiered control over cloud resources can be. I was especially pleased that RightScale incorporated our suggestions for improved sub-account control into their development cycle," continued Dr. Fusaro.

The palaver project teams also leveraged RightScale's unique approach of managing complete cloud deployments, in which each server can be pre-configured and controlled using a cloud-ready ServerTemplate. A ServerTemplate starts with a "RightImage," which is a simple base machine image that normally contains only the operating system, and then adds scripts that define the role and behavior of that particular server. These "RightScripts" may run during the boot, operational, and shutdown phases of the server's lifecycle.

The use of RightScripts allows for easy and persistent configuration of environments. Mr. Prasad Patil, a co-administrator of the LPM cloud management team, implemented a RightScript to mount an EBS volume full of data and executables and install software packages onto an extra-large, eight-core instance. This provided him the opportunity to shut down his otherwise expensive instance between analysis runs and pick up where he had left off upon relaunch.

Dr. Fusaro and Mr. Patil coached the other groups to create similar setups, allowing for a fully configured development environment upon launch of the instance. The deployment methodology and organization created tracking of what was being launched and by who, which saved time on inventory runs.

With the conclusion of the Harvard palaver, the team confirmed the choice to use RightScale expedited the processes of resource management and development of best practices at Harvard Medical School. With RightScale, the project teams reduced the learning curve for cloud resource administration and greatly improved collaboration among distributed project teams.

"Cloud computing managed by RightScale is a fundamental resource for scientific research. We accomplished our goal to significantly reduce overall administrative and management requirements," concluded Dr. Tonellato.