Designing fault-tolerant distributed archives for picture archiving and communication systems

Purpose: Distributed archives in a picture archiving and communication system (PACS) environment can provide added fault tolerance and fail-over capability, as well as increased load capacity at a more economical price than traditional “high-availability” systems. Systems can be configured with varying levels of fault tolerance, depending on the amountof redundancy desired. There is, however, a direct correlation between the level of hardware redundancy and cost to implement. This presentation details the system design for fault-tolerant distributed archives as well as several options for redundancy, referencing implementation of a fault-tolerant archivesystem at the University of Utah.Methods: The distributed archive system described here is based on Image Devices’ image archive software, which can be implemented on multiple individual archive servers in order to distribute archive functionality and operational load. The configuration and implementation of the individual servers together make up the distributed archive system and does not impact the ability of the system to be scaled to meet future requirements. Several implementation and configuration options exist, including the ability for servers to maintain replicated databases containing pateintand image information. Thus, each archive can be aware of all information and the location of this information within the distributed archive system.Results: The goal is to produce systems that will still be operational in the event of any single point of failure, ie, a network connection failure between facilities or the failure of asingle archive server within the distributed system. During normal operation, workload forimage acquisition, image routing and image query requests will be distributed between the archive servers. If the system is deployed in a multifacility environment, each archive server can be configured to be responsible for the acquisition and image distribution management within that server’s localfacility. If the system is deployed in a single facility environment, load can be distributed evenly between the archive servers based on an understanding of the workload requirements generated be each acquisition and display device in the system. In the event that an archive server fails, other archive servers within the system will have the ability to provide redundancy employed. Three levels of fault-tolerant design can be achieved with this system architecture: (1) duplicate work capability only; (2) duplicate work capability and short-term image cache; (3) duplicate work capability,short-term image cache, and long-term image archival. Using the basic fault-tolerant design above, we have implemented a multifacility distributedarchive system at the University of Utah. This system was implemented at a fraction of the cost of true “high-availability” archive architectures yet provides constant up time for the PACS system. If the network connection between thetwo locations goes down, each siteis still fully functional for soft-copy read, as well as image acquisition and distribution. If either of the archiveservers goes down, the image sources are redirected to the other archive server. The operational server then handles image distribution for both locations. Access to images in the short-term image cache is available to both archive servers and is not affected by loss of the network connection or remoteserver. Because there is ony one long-term archivedevice, the ability to retrieve images from long-term storage is theonly function compromised by a network or server failure.Conclusion: By implementing distributed archives in a PACS environment, it is possible to achieve a highly fault-tolerant system without the expense of high-availability hardware and software. The design concepts outlined here can be applied to any PACS system that supports distributed archive functionality.