shadow

Choosing the right solution

Implementation

There are two principal options for deploying CRAB – either as a confined system within a hospital or as a remotely hosted service, fully encrypted and accessed securely through a certified N3 connection. Either option guarantees data protection and is CfH compliant as a minimum standard. See below.

Internal Deployment

Deploying to Trust servers is a sensible option for an organisation to feel assured that it has full control of it’s internal data and that information integrity and confidentiality is guaranteed. This is readily possible provided that the hospital runs a Microsoft SQL 2005 database and an IIS 6.0 web server.  In such cases, deployment is straightforward and inexpensive.  This has proved to be the case in NHS hospitals where SQL 2005 clustered servers and intranet servers are becoming standard.  Alternatively, instead of deployment to the hospital’s existing servers, CRAB could be supplied as a stand-alone server requiring simple connection to the hospital network.  However, because of the hardware and Microsoft licensing costs, this is a more expensive option.  Although CRAB uses straightforward standard Windows technology, minor training of the IT staff in each hospital would be required.  CRAB Informatics offers specialist remote assistance to IT personnel, limiting the day to day local requirement to routine maintenance and monitoring. 

External Deployment

Running CRAB as an external service is also possible. It has the advantage that there is no significant IT support requirement in each hospital; it is straightforward to implement site to site comparisons and, subject to authorisation, reports may be viewed from any location.

The external service can be run by CRAB Informatics at no additional cost to the standard licence fee.  This option has been trialled at a private provider site with secure access enabled across the internet.  There were no interruptions of service during the trial period.

The architecture that was used for trial is capable of supporting a series of hospitals simultaneously and could be extended using standard Microsoft architecture to any number of other clients without interruptions to service or breach of data integrity. 

Provisioning CRAB across the internet requires measures to protect against loss of a single hosting centre or its connectivity.  During the trial the primary hosting location went off line for four hours due to the failure of a major network switch.  Service continued through this period from the secondary location with no loss of data.

Two identical CRAB server deployments are located in separate hosting centres to ensure availability in the event of failure of one server or the hosting centre or its connectivity.  Mirroring software is used to clone the primary location to the secondary location every two minutes.  The primary server is checked continually from three separate locations and if failure is detected by two or more monitoring servers the internet address of the CRAB URL is changed to point to the secondary location.   The service then continues from the secondary location.  Once the fault at the primary location is cleared, the servers are resynchronised and the CRAB URL again pointed to the primary location.  There are many ways this could be achieved but this architecture allowed robust operation at low expense.  By using RAID servers, redundant power supplies and uninterruptible power feed, each location is intrinsically robust. 

In a production environment, a real-time data mirror could be employed to ensure rapid failover and immediate recovery.  This would require a high speed internet link between the hosting centres which is costly.  It may be decided that a high availability deployment (clustered database and web servers) in a single hosting centre with multiple network connections would provide adequate performance.  In a separate product deployment on clustered equipment in Harbour Exchange, Docklands, London, one of the United Kingdom’s most prestigious hosting centres there were two outages of more than four hours in an eighteen month period.

 

quote