.

AWS Implementation for Quantitative Digital Pathology at Scale
26 November 2020
By

The Challenge

To meet increasing demand from Clients, OracleBio recently expanded its image analysis software capabilities to include increased licenses for the Visiopharm platform including AI capabilities, as well as an integration of Indica Labs’ AI module to its existing HALO platform. As a result, the existing IT infrastructure required reconfiguration to address the increase in storage and computing power needed to take full advantage of the new software. One option was to expand on the current on-premise infrastructure. However, it was determined that a cloud computing solution would provide a more flexible and scalable infrastructure with on demand access to GPUs and CPUs required for AI training and image processing.

OracleBio required a scalable solution that enabled its team of scientists to be working on multiple studies at the same time. Image files processed by OracleBio can be very large, often in excess of 50Gb each, so the company needed a solution that could not only process such large files easily but also provide storage for these files and associated data after study completion. The nature of OracleBio’s work means that data protection is of utmost importance, so the availability of comprehensive security and encryption features was essential.

    The Solution

    There were several key areas that OracleBio was keen to address in order to provide a suitable solution.

    • Computing compatibility – The company already had a substantial investment in on-premise equipment, including a modern GPU enabled server. This equipment still has a useful working life left and so any solution had to be compatible with existing resources.
    • Fast parallel processing – Increases in the number and the size of images to be processed created an increased demand for parallel processing, rather than sequential image processing on a single server. The speed of batch processing of images on the existing on-premise hardware was restricted by available server capacity. The company also needs to be able to respond quickly to changing customer requirements on a study-by-study basis. A hybrid solution, comprising both on-premise and the AWS platforms, that allowed capacity to scale up on-demand could materially reduce turn-around times for large studies.
    • Meeting increasing workflow demands – As the company grows and with increasing customer demands a one-size-fits all infrastructure solution could not meet the needs of the company for all types of studies.

    Broadband Cloud Solutions (BCS) was able to assist by application of their extensive knowledge of various Amazon Web Services (AWS) products and services. Key AWS services that are relevant to OracleBio include:

    Computing Power: The software products used by OracleBio require substantial computing power, often involving Graphics Processing Units (GPUs). AWS Elastic Cloud Computing (EC2) allows customers to set up computing requirements virtually and on-demand, eliminating the need for any capital expenditure on, otherwise expensive, infrastructure.

    Storage: The source data and processed images require large quantities of data storage. Amazon’s Simple Storage System (S3) allows OracleBio to encrypt and securely store the relevant data. S3 provides different levels of storage to  cover regularly accessed data, infrequently accessed data and long-term archived data. Automatic methods manage the storage lifecycle to optimise the cost of storage for OracleBio.

    Security: The OracleBio solution takes full advantage of Amazon’s security features. All data is encrypted at rest and in transit and securely stored within the UK. Network security is in place at three levels and access to the working environment is secured by VPN.

    Implementation

    OracleBio established user requirements and specifications in close partnership with the BCS AWS Certified Solutions Architects (SA). An in-depth review, by BCS, of OracleBio’s vision and existing landscape enabled potential solutions to be designed. This in turn enabled OracleBio to gain a more in depth knowledge of AWS and its future potential for the organization.

    Testing of the compute and storage services of AWS and especially the GPU enabled instances, required for deep learning AI processing, was carried out as part of an initial feasibility study. The feasibility study was such as success that OracleBio embraced the platform enthusiastically and production workloads were established in the cloud within days.

    The final solution meets OracleBio’s needs and is based on Amazon Web Services’ ‘Well-Architected Framework’ (WAF). This utilises operational practices across five main pillars:

    • Security
    • Cost Optimisation
    • Performance Efficiency
    • Operational Excellence
    • Reliability

    The nature of data hosted by OracleBio requires the company to agree to high levels of Security and Operational Excellence and so part of the BCS solution was to deliver a high standard of documentation to satisfy the requirements of customers.

    And the effect for OracleBio’s clients?

    Perhaps the biggest advantage in using AWS for OracleBio and its customers is the business agility aspect.

    • Image analysis study scientists are able to request and utilise required infrastructure and computing power on studies, which can be provisioned within minutes. This provides OracleBio’s clients access to vast computing infrastructure whenever required, without the cost implications of on-premise hardware. This provides a tremendous advantage when there is a need to deploy multiple analysts on urgent image analysis studies.
    • Parallel processing enables fast processing and significantly improves turn-around times on Clients studies.
    • The AWS ‘fail fast’ reporting interface enables flexible testing and product application development which was not possible previously with on-premise hardware. This facilitates OracleBio’s development of innovative computing solutions to satisfy client needs, such as OracleBio’s OBserver image viewing platform.

    Find out more about our Software Capabilities

    OracleBio's Quantitative Digital Pathology Services

    Stay up to date with OracleBio

    Sign up to our mailing list to stay informed about news, webinars, events and more