Wyeth is a global provider of prescription pharmaceuticals, non-prescription consumer healthcare products and pharmaceuticals for animal health. With items sold in more than 140 countries, and 52,000 employees worldwide, the company netted revenues of $17.4 billion in 2004. Headquartered in Madison, New Jersey, Wyeth reinvests almost 15 percent of revenues into research and development initiatives to create innovative pharmaceutical, biotech and vaccine technologies. Situated within the company's research division in Cambridge, Massachusetts, is an organization chartered with harnessing computers to further the understanding of biology -- Wyeth Bioinformatics.
John C. Morris, manager and principal engineer within Wyeth Bioinformatics, explained the terminology, " 'Bioinformatics' describes the scientific use of computers and algorithms to model and analyze vast amounts of biological information obtained from internal research, public databases and commercial data. Specifically, our focus is on protein, RNA, and DNA based scientific systems. Our small but skilled team has over 35 years of combined experience deploying computational technology to enable the critical needs of research to analyze biological information. Wyeth Bioinformatics is responsible for the acquisition, support, administration, and analysis of the massive amount of bioinformatics data that is generated throughout Wyeth."
Bioinformatics gets attention
Recent years have seen extraordinary growth in the field of bioinformatics: Completion of several high-profile initiatives, including the Human Genome Project, have raised awareness and fueled interest in the discipline. In addition, new technologies such as high content cell screening are adding to the data and analysis needs. With this upsurge has come an associated explosion in the volumes of data and need for powerful computational resources. Morris elaborated, "Due to the growth of existing systems and additions of new scientific technology, we typically see data volumes doubling every 12 to 18 months, and this in turn places a perpetually escalating burden on our processing capabilities."
Morris further described the challenge that he was facing, "When we coupled the increasing costs to support and maintain our existing legacy systems with a diminishing ability to satisfactorily service the research teams' computing needs to analyze this growing amount of data, we knew that changes were needed."
Morris and his team developed a vision and began to craft a set of requirements for a brand-new environment that would be a catalyst for bioinformatics innovation, and at the same time fulfill his own stipulations for security, stability, continuity, resource allocation, and performance. His analysis lead to the identification of key infrastructure components -- a clustered Intel-based configuration, open source operating systems such as Linux, and specifications for critical applications for networking, storage, job scheduling and resource management.
Morris described his thinking, "We wanted to increase our productivity and compute capabilities, while simultaneously decreasing costs and administration. The architecture had to easily accommodate the latest, and best of breed technologies coming from partners, commercial vendors and academia. We felt that Linux would facilitate this, as well as lower our capital and operating costs. The free-market forces that surround Linux and the many open source initiatives were of great attraction to us, and will ensure that the momentum for innovation and cost containment continues. Additionally, the architecture had to be secure, highly available, able to dynamically allocate resources, and capable of disassociating applications from specific hardware -- we determined that a clustered approach would fulfill all of these requirements."
The hunt begins for the perfect infrastructure
Wyeth embarked on a swift two-month initiative that involved creating a request for proposal (RFP) based on the desired hardware, software and services characteristics and circulating it to a number of different vendors. A short-list of three finalists was made, with each being required to conduct further presentations and on-site demonstrations of the proposed solutions.
Morris described the outcome, "We chose HP. Although we currently use HP ProLiant servers for our Microsoft Windows applications, and had a high level of familiarity with HP's products, we performed the evaluation using only the criteria defined in the RFP. We objectively assessed all of the responses, and HP's solution had the best value. We especially liked the integrated 'one stop shop' approach that HP offered -- this really set it apart from competing proposals."
He continued, "We even acquired our Red Hat Linux licenses through HP -- the indemnification that they provide for the product allowed us to fully mitigate any potential risks and remove any unnecessary distractions associated with the decision. The HP architecture is equally at home with either SuSE or Red Hat versions of Linux, but the Red Hat variant is the operating system of choice for many of the academically-oriented applications in which we are interested."
The selected cluster configuration consisted of multiple HP ProLiant BL20 Intel-based, dual-processor server blades and several ProLiant 380 servers for administrative, management and logging tasks. HP ProCurve Switches interconnect blades with each other, as well as to the storage environment.
Morris commented, "We especially like the cluster's inherent ability to eliminate any single point of failure. Given the run-times, number, and frequency of some of our calculations, and the globally diverse user community, any failure has severe implications. With a cluster, even if multiple compute components fail, the overall capabilities of the system are maintained. The scientists can continue working and do not have to worry or modify their activities like they may have in the past."
To ensure a smooth deployment, HP delivered a blend of consulting services that included initial configuration and setup, implementation, and training -- covering both operational and end-user topics. To take full advantage of the capabilities of HP's high-performance cluster, Wyeth purchased LSF HPC scheduling software from Platform Computing. By optimizing throughput and minimizing management and communication overheads, all resources can be operated at maximum capacity.
The results are dramatic
Morris enthused about the results, "We went live in September of 2004, and have seen a dramatic decrease in our capital and operating costs -- we had payback in well under 12 months. In general terms, we increased our computational capability 40-fold -- with our vastly increased processing power and decreased costs, we definitely got value for money and achieved our corporate mandate for increased productivity and innovation!"
Another benefit has been that the limitations imposed by the legacy environment that caused researchers to adversely alter their use of the equipment have been removed. Morris explained, "Previously scientists would perform a less stringent first-pass analysis to reduce the size of their data sets, and then use the more computationally intense transactions only on the reduced data. Now they are able to use more rigorous analytical procedures from the outset and therefore have increased the accuracy and statistical relevance of their overall results. It really liberated us from having to do so much pre-processing of data and incurring the risk of losing important information."
He continued, "The environment provides discovery scientists with the capabilities to do analysis that we just couldn't do in the past. Work is completed faster and the quality is better -- more frequent and updated data samples, better statistics from larger data sets, a better electronic laboratory space, and more scientific explorations were enabled. We've been able to push Bioinformatics into domains that we previously hadn't considered. Because we have removed some of the original barriers, people are finding that this is a tremendous asset for them to use in furthering their research. We have customers that we never anticipated, and we have uses that came about from new innovations. It is imperative to have this dynamic capability in research -- the strategy we chose, the architecture we implemented, and Linux allow us to deliver."
Morris summarized, "This project represented a very significant change for us, and because we put so much emphasis into ensuring that the infrastructure would exactly meet our research requirements, we had very high expectations -- HP has done really well in meeting all of these. We wanted to lower costs, increase capabilities, migrate terabytes of data, move in excess of a hundred highly demanding users and applications, and get payback within 12 months -- all extremely aggressive goals -- we accomplished them with the assistance of HP's implementation and training support, and excellent Linux server price/performance."
-----
Source: Hewlett-Packard
More info on HP’s Life and Materials Sciences More info on HP’s HPC solutions |
Printable version
