OPTIMIZED SYSTEMS

Healthcare Tackles Big Data

Michael Steinhart | Date: 02-16-12 | Comments

Workload-optimized systems for healthcare ensure that each application has the optimal mix of resources for high performance, reliability, and efficiency.

The healthcare industry generates enormous amounts of data every single day. If you�ve ever had a CT or MRI scan done, you know that each one generates dozen�and sometimes hundreds�of high-resolution images. A busy medical center can see several hundred patients per day, generating hundreds of gigabytes of data that must be analyzed, backed up, and archived in long-term storage.

Electronic health record systems, or EHRs, are also pouring terabytes of information into medical databases around the country. Patient histories, test results, prescriptions, and treatment records are all combined into EHRs that can be shared more easily among healthcare providers and analyzed by powerful bio-informatics applications.

Genomics is another field that is pushing the boundaries of storage and analytics. In 2000, the first genome sequence took up about 750 gigabytes of space. Today, increasingly sophisticated technologies at research institutions, universities, pharmaceutical companies, and even hospitals are generating terabytes of genetic information each day. At Cornell University, for example, the computational biology service unit collects about a terabyte a day from its affiliates. This information isn�t sitting in an archive, either: It�s being crunched by supercomputers in near-real time.

Right now, you need a supercomputer � or a cluster of high-performance systems � to conduct bio-research with such huge amounts of data. But workload-optimized systems can help doctors leverage this data to improve everyday patient care, too.

As we�ve discusses in the past, workload-optimized systems involve buying and deploying the technology that�s best-suited for your workloads, and ensuring that each application has the optimal mix of resources for high performance, reliability, and efficiency.

Some applications may need large amounts of memory, for example, and others read data but don�t write any. The mix of hardware should correspond to these specific usage patterns, instead of just buying the latest and greatest and assuming it will handle every workload.

Hospitals can benefit from workload-optimized systems by devoting high-performance resources to data analytics and more mainstream equipment to administrative tasks, for example.

In fact, eWEEK reported about medical applications that are being developed by health insurance firm WellPoint for IBM�s Watson supercomputer. The hardware is tailored�or optimized�for WellPoint�s needs, and the applications help doctors and nurses with diagnostics and treatment plans.

Using its data analytics and NLP capabilities, Watson would integrate data such as , patient histories, , and to help doctors decide on courses of treatment.

Using Watson or similar analytics engines, healthcare providers can analyze patient care histories, test results, recent research, medical literature, clinical trials, outcomes data, drug information, and potential side effects. This can help prevent drug interaction problems and other diagnostic mistakes, which account for 20 percent of medical errors, according to Herbert Chase, Professor of Clinical Medicine at Columbia University.

Chase explains that biomedical literature doubles in size every seven years, and it�s much more information than any doctor or any single computer can store and process. These new apps don�t replace doctors, of course, but they help provide a more comprehensive resource for analyzing more medically relevant information than ever before.

The bottom line is that healthcare is ideal for workload-optimized systems, because they can help provide faster access to personalized medicine�resulting in improved patient care. In 2012, vendors will continue to develop cloud databases and supercomputers�such as IBM Watson�to store and process large volumes of "big data" collected via EHR and genomic data stores.