Design of a Scalable and Fault Tolerant by kzp12233


									              Design of a Scalable and Fault Tolerant
           Enterprise Medical Image Management System
                                       Chris Hafey
                                       Stentor, Inc.
                                 South San Francisco, CA

 Over the past decade, health care organizations have been looking to reduce costs and
improve patient care by utilizing computer technology instead of film. Today there are
software computer based solutions that enable automated workflow, rapid enterprise-
wide image distribution, long term archival of acquired images and software for
improved image interpretation by radiologists. Collectively, these solutions are known as
a PACS (Picture Archival and Communication Systems).
 Despite a variety of technical solutions to the PACS problem, no one has been able to
deliver an acceptable solution. The challenges facing a PACS system are 1) The medical
imaging data sets are too large for standard networks which prevents the system from
scaling to enterprise use, 2) the hardware cost to provide redundancy has been too high to
provide a fault tolerant archival system, 3) the standardized medical image protocol
(DICOM) has no provisions for fault tolerance, 4) the image data sets are too large for
standard PCs to manipulate in real-time.
 These challenges have given PACS a bad name – they are known for being slow,
unreliable and expensive. The best PACS solutions can approach 98% uptime, but this is
not typical – some systems are down for weeks at a time! Considering that 80% of
hospital visits result in a radiology procedure, this kind of uptime is not acceptable to any
organization (including HMOs ).
 Stentor is a 3.5-year-old startup company that was founded to solve enterprise wide
image distribution using wavelet technology for image representation. Our initial product
iSite Enterprise 1.1 was designed to run on a single server and scale to 10 simultaneous
users. The product has been a tremendous success resulting in over 100 customers and
one site pushing it to handle 150 concurrent users! Due to the product’s speed,
intuitiveness and reliability, customers began asking us for a complete solution to replace
their existing PACS. This prompted us to develop our next generation product to
overcome the challenges our competitors have not been able to solve. This next
generation product is named iSite 3.0 and is currently deployed at 12 sites nationwide.
 To achieve 1) 99.99% uptime and 2) scalability from low volume community hospitals
to the largest health care organizations, this presentation will describe the following key
features of the systems design and architecture: 1) Fault tolerance provided by the
hardware, 2) Fault tolerance provided by the software, 3) the different deployment
configurations designed to meet the different site volumes. This presentation will also
focus on how TAO was utilized in the software to implement the fault tolerance and
scalability features as well as the lessons learned from real world deployments.

To top