Amazon Web Services Are Coming!
- DGIT Secure Research Computing Platform. All services running in AWS.
While primarily known as a bookseller, online marketplace, and media company – all of which are true – Amazon is also involved in a growing assemblage of additional industries. One of which, Amazon Web Services (AWS), comprises a series of more than seventy distinct, on-demand, cloud-based computing systems. While the sheer size of AWS – a $14 billion business on its own, around since 2007 – is less revelatory than it was given the rising prevalence of cloud-computing, it perhaps remains surprising that AWS is so critical for Amazon’s profitability, accounting for only 7% of their revenue, but 71% of Amazon’s operating income.
Widespread migrations to cloud services, particularly in healthcare, are accelerating. Upending traditional high-volume procurement approaches once-required to manage uneven demand, cloud business models allow IT organizations to right-size and scale on-demand. Services are used when needed, and disabled when not. As customers, we pay only for what we use. In an oversimplified example, a typical server, traditionally-procured and managed, costs UCLA $15,000 over 5 years. A similar server in AWS costs DGIT less than $2,000 over the same period, assuming the costs are fixed. In a highly competitive market, AWS costs have dropped every year of its existence.
DGIT has big plans for AWS, starting with research computing. Today’s bioinformatics research requires secure, cost-effective, scalable technologies that enable prevailing big-data, machine learning, and artificial intelligence capabilities. The designation of the Institute for Precision Health (IPH), the launch of the Depression Grand Challenge (DPC), efforts to rapidly advance drug discovery and create new avenues of intellectual property development, deep integration of genetic testing and analysis into clinical care, and many other significant and potentially transformational efforts in UCLA Health, demand instant access to computational toolsets. These tools are readily available, but for security concerns, cost apprehension, or general lack of awareness, are not widely used by DGSOM. Developed in concert with expert UCLA faculty, DGIT will provide secure, cost-optimized, scalable services that enable new insights, augment discovery, and accelerate the next generation of breakthrough innovations.
Initial efforts include developing the underlying security required to achieve HIPAA compliance within AWS. Once those are complete (expected by June 2017), creation of high-performance computing capabilities, used for genomic sequence analysis and other computationally-intensive workloads, will begin. In conjunction, DGIT is creating a hub for cohort discovery within AWS. Made up of a genomics database allowing for fast querying of millions of gene sequences, a complete copy of the UCLA electronic medical record, purpose-built for research, and query tools starting with an updated version of i2b2, this hub will provide researchers with on-demand access to patient-based data sets correlated with genomic data, and facilitate searches for research cohorts and related data sets (e.g., display a de-identified list of all patients with specific genotypic and phenotypic qualities and output de-identified EMR notes from each patient record).
Through this platform, researchers will access de-identified data on-demand, and identified data based on institutional review board (IRB) approval. As cohort data is collected, researchers can apply AWS-based computational toolsets, including high-performance computing, machine learning, artificial intelligence, and natural language processing, to aid in discovery efforts. Custom or workload-specific applications (e.g., MATLAB and R) can be launched via DGIT-managed systems or implemented directly by researchers within the secure platform.
DGIT is working on cost models and usage options with the goals of keeping the platform costs under-market and available to as broad a swatch of the UCLA community as possible. Please contact us if you have interest or questions in these or other research computing services. We look forward to partnering with you.