Opening up grid technology

Headquartered in Chicago, the OGF is a non-profit consortium of business, scientific and academic organizations and individuals dedicated to accelerating the pervasive adoption of grids worldwide. NGP spoke to President and CEO Mark Linesch to get his thoughts on how pharma is at the leading edge of grid technology

NGP. The pharma industry is at the leading edge of grid technology but many companies still have large computational and data integration issues. What evidence have you seen of this?

ML. Being at the leading edge is often described as the “bleeding edge” – it clearly comes with a set of opportunities and challenges. The opportunity is competitive advantage and the challenges are managing the people, process and technology risks effectively.

From a technology perspective, the pharma industry is dealing with an explosion of data that must be integrated, processed, analysed and shared – particularly with recent developments in genomics, proteomics and emerging areas such as systems biology. Grid technology helps deal with the computational and data integration challenges and enables researchers to collaborate more effectively but the technology is still maturing. Even more significant may be the non-technology, people and process issues. The drug discovery process is increasingly a digital process that is replacing and/or complimenting traditional experiments. Taking the scientist or researcher from the familiarity of the lab to the new tools of the grid introduces cultural and education issues. Developing the organizational and security policies for drug discovery with external partners takes thoughtful consideration. Accelerating the overall life cycle introduces opportunities to re-think established processes with an eye toward both incremental and dramatic improvements. Successful organizations seem to focus first on enabling a small number of users and applications and then building on this success rather than the “build it and they will come” approach.

NGP. In a recent article you said the GGF sees the evolution of grids from an academic curiosity to a commercial technology as happening in stages. Where does pharma fit in?

ML. I like to first view the evolution of grids within the context of the broader industry trends. In today’s dynamic business environment, enterprises are aggressively working across organizational and geographic boundaries to shorten lifecycles, deliver greater customer value and competitive advantage. Working horizontally places increased demands on organizations to break down existing barriers that inhibit the flow of information, innovation and commerce in our increasingly inter-connected economy. Individuals are challenged to work in new ways – often in collaboration with other departments, disciplines and/or organizations. IT is challenged to deliver increasing value to the organization while reducing costs and maintaining stability as they transform their organizations to support this more dynamic, global business environment. Solution providers are challenged to deliver practical real-world solutions that deliver value today while helping to cut through jargon and provide direction on future IT investments

Within this overall business environment, I think of the adoption model for grid in three phases: (1) early adoption; (2) proven solutions and (3) pervasive adoption. Early adoption is primarily an exercise in handcrafting solutions. In the proven-solutions phase, you see grid-enabled software from vendors and grid success stories in specific industry sectors. These proven solutions provide real-world examples of the benefits and risks of grid deployment and enable other organizations to leverage the successful experiences of the early pioneers. In the pervasive-adoption phase, mainstream users can start to adopt grids with packaged solutions. Moving to more pervasive adoption requires the “lessons learned” from early adoption, the “success patterns” from proven solutions and breaking through the non-standard barriers so that resource sharing and collaboration can be more dynamic and extend beyond the enterprise the broader health care community. We’re getting there, but it takes time.

My impression is that pharma has been one of the pioneers in applying grid technology to compute and data-intensive problems within their industry. Grids and next generation distributed computing technologies have great potential to accelerate drug discovery, enhance patient safety and fine-tune therapeutics. A recent study by Robert Cohen from the Economic Strategy Institute indicated that adoption within the pharmaceutical pndustry would occur in stages with wide-spread adoption occurring toward the end of the decade.

NGP. You are also quoted as saying if you want to get broad adoption and pervasive grid-like qualities you need standardization and interoperability. How has this been progressing?

ML. Regarding how standards and interoperability are progressing, I think the need is becoming more critical as organizations progress beyond their initial grid deployments and they extend to other part of the enterprise or between trusted partners.

Today, organizations can adopt grid solutions and achieve significant benefits even without having all the standards in place for interoperability. This is particularly true for enterprises running the same version of one of the popular grid middleware software products. Interoperability becomes more critical as organizations connect grids to other grids within their organizations or with other organizations that utilize different grid middleware software. For those grids to come together and interoperate, they need to speak the same language.

As the grid solutions become more widely adopted and the need for interoperability increases, the entire industry has a role in creating a clearer sense of direction and urgency in accelerating grid and distributed computing standards adoption. OGF must focus our standardization efforts and clearly articulate practical, near term priorities. We must communicate more clearly and collaborate more effectively while obtaining active participation from key industry experts. The broader distributed computing community must continue to make progress on the foundational Web services standards upon which many of our specifications are based. The vendor and open source community must productively engage in the open standards process and adopt the specifications that enable software interoperability. Finally, end users need to encourage vendors to deliver software based on industry standards and to provide support for new software licensing models. It’s quite a task and OGF cannot do this alone – it does take a “community” environment.

NGP. What particular standards are organisations such as the OGF backing and how close are they coming to fruition?

ML. OGF architectures and specifications are designed to align with existing standards and emerging web services specifications being developed within the broad computer industry. This allows grid practitioners to leverage the tools, educational materials, and experience when building and managing grids. This broader industry process has included significant work to establish and agree upon a set of web services standards (SOAP, XML, WSDL etc.) that enable interoperability within a distributed computing context. The industry is currently working to build on these core standards and extend the web services paradigm to the management of grid resources and events through a set of converged web service specifications now under development.

OGF has developed the Open Grid Services Architecture (OGSA) and continues to mature the related grid specifications that build upon these broad industry standards to deliver a set of foundational capabilities for interoperable grids. These capabilities include specifications for discovering, monitoring and controlling resources; describing, executing and managing application services on these resources; managing data access and data movement; and insuring secure authorization and access privileges for those utilizing the grid.

It is anticipated that the combination of existing standards, newly converged web services specifications and OGSA-related specifications will enable progress on interoperability over the next 12 to 18 months.

NGP. Some software being used is grid compatible straight off the shelf but some still has to be adapted. What are the main challenges in this area?

ML. I think there are several challenges to running applications on a grid-like infrastructure. The first challenge to running applications on a grid-like environment is to determine whether the application is suitable. Applications that are modular and can execute as a series of steps within a broader orchestrated process are often ideal. You also need to consider compute/data and latency dependencies – particularly when executing tasks within a loosely coupled, geographically distributed environment.

The second challenge is often tuning the application to run efficiently on a grid-like infrastructure. Many of the early grid successes have been with High Performance Computing (HPC) applications – either developed internally or purchased from an HPC software vendor. Many of these applications already support scale-out cluster and grid environments or can quickly be adapted to work in a grid environment – often in a few days or a few weeks. Traditional enterprise application vendors such as SAS, Oracle and SAP are also gaining experience and beginning to support grid environments although this functionality is often implemented with consulting services to ensure success. Finally, newly develop service-oriented applications are designed to work in shared, grid-like environments.

The third application challenge has to do with the current ambiguities regarding software licensing. In the case of licensing, while there are a few “technical” issues, most of the barriers involve non-technical, economic and policy issues that the software industry is currently sorting through. In the meantime, individual companies are working with their software providers to articulate requirements and determine both short and longer term approaches to licensing software on grid-like infrastructures.

NGP. A lot of pharma companies are running a relatively small number of applications on the grid. What potential does the grid hold and how can firms exploit this further?

ML. Moving toward grid-like infrastructures and automating and integrating drug discovery, trial, and deliver processes is clearly a journey. The potential for grids in pharma is exciting but should be based on a practical, measured, approach – this is a marathon, not a sprint.

Organizations have the opportunity to utilize grids to shorten and reduce the cost of developing new “blockbuster” drugs that now cost $1 billion to develop (recent Tufts study). Grids enable organizations to run increasingly complex applications that require ever larger datasets and multi-step analysis. They enable scientists within a single organization to have access to the company’s research data and/or enable researchers from different organizations to collaborate on a single drug development effort. Grids also create ways for people or companies to contribute computational power to the study of diseases that we know little about (for instance, the Intel collaboration with Cambridge University to examine anthrax).

I often say that grids are a core part of the next stage of distributed computing, and it’s important that they look for specific problems that they might solve with grid today. Then start small. Let the infectious nature of grids lead to additional applications and solutions.

Johnson and Johnson is an example of a company that started in 2004 with a small number of applications and a focus on the user. By the end of 2006, they have indicated that they could have as many as 26 applications running on their grid. I think they have done a great job managing the hype and focusing on the value – letting “the infectious nature of grids” take hold in a practical, measured way.

NGP. With so much secrecy and competition in the pharma industry, IT and computer experts may still be wary of scavenging or loss of control when it comes to data gathering. What evidence is there to allay these fears – is better training or more information needed?

ML. I would say that more information and better training is needed but I don’t know that the fears that you mentioned are unique to the pharma industry. The scientific and engineering process within many industries is now a “team sport” and it is not unusual for competitors to also collaborate on a costly research project or design. In the information age, data is acknowledged as a key corporate asset. As data has become more central to the creation of value, the successful management of trials, regulatory compliance, and integrated patient care; organizations have instituted policies to insure that sensitive data is kept secure. Most pharmas keep a master copy of any database that they use and employ security software to identify who has accessed a database or has attempted to modify it. When data is shared between large pharmas or between several smaller pharmas and a large one, there are strict procedures in place to insure that there is no scavenging or loss of control. Much time and money has been spent to secure data that is critical to developing new drugs.

I am not sure there are many case studies that reveal these steps because pharmas are reticent about discussing security measures. It might be helpful for them to be a bit more forthcoming about some of these steps because it might allay fears that experimental data can be easily stolen.

NGP. What other advancements would you like to see when it comes to grid technology and standards that could lead to more widespread grid use, and how are they being realised?

ML. I am intrigued with the opportunities for horizontal integration – enabling information/resource sharing and collaboration across what we today perceive as difficult data integration and/or organizational boundaries. Grids and grid-related technologies such as virtualization, automation and service-orientation will mature over the next few years as organizations gain experience and vendors incorporate the specifications into their products and develop the tooling for more widespread adoption.

As grid technologies and standards continue to mature, I particularly like the ability of grids to bring concepts such as personalized medicine into focus. I was in Taiwan recently where collaboration between the government, medical and commercial organizations is delivering better care to asthma patients and better information for improved treatment and prevention. Using an advanced data grid and an interface to asthma patients through wireless phone technology, the physician is able to engage with patients in a more real-time manner. Bringing together weather, environmental, drug and protocol information enables better patient care immediately but also has the potential to provide valuable information exchange with other public and private health care organisations for future disease prevention and improved drub and treatment alternatives.