Purpose
There are many mission critical systems in many industries which are candidates for High Availability implementation. High Availability implementations minimize the downtime for these systems. Interface Engines, Cloverleaf, eGate, Datagate, Jcaps, etc., are a prime example especially when they are used in healthcare. Most data transmissions in a hospital pass through an interface engine so downtime impacts patient care, workflows and efficient operation when staff must revert to manual procedures. A High Availability implementation decreases downtime to a minimum.
There are typically two major models for implementation: Active-Passive or Active-Active. While both minimize downtime, there are strong points for using an Active-Active model.
High Availability Implementation
High Availability (HA) implementations consist of multiple servers configured together into a single HA cluster. The servers communicate with each other to update status and maintain a heartbeat. The servers host applications while the application data and configuration are store on a SAN with RAID disks. If there is a hardware or network failover that interrupts the normal operation of the application(s) hosted on a server, the application(s) are transferred from one server to another. The amount time required to transfer the application(s) is a function of the application(s) orderly shutdown and restart times.
An HA implementation consists of the HA vendor software and the application scripts it invokes to startup and shutdown the applications when a problem is detected. The applications will typically have support and maintenance scripts which will be executed on a periodic basis.
Active –Passive Model
Typically, mission critical systems have production and test systems. Additionally, development systems may be present as well when test systems are heavily utilized for training and integration testing for system upgrades.
In the Active-Passive model there are two servers in the Cluster. The primary server is active and hosting the production applications as well as test applications and optionally development applications. The secondary server during normal operation does not host any of the applications and is essentially dormant just running the operating system.
This configuration is easy to implement than the Active-Active Model since there is only one set of start and stop scripts. Management of the scheduled support and maintenance scripts is straightforward as well.
The major downside of the Active-Passive Model is there is nothing operational on the secondary server. There is no guarantee the production applications will function as expected on the secondary node due to ever changing networks and servers being hosted on the networks. Also expensive hardware is not being utilized.
A derivation of the Active-Passive Model has the test and development applications hosted on the secondary server but they are not configured for high availability. With applications running on the secondary server that is functioning and communicating with other test and development servers, there is a very high probability the production applications will properly function if a fail over from the primary server occurs. The test and development applications may be shut down after a failure over-dependent on the implementation and the HA application scripts. A downside of not implementing the test and development applications in HA is downtime when performing maintenance or upgrades on the secondary server.
Active-Active Model
In the Active-Active Model the test and development applications are hosted on the secondary server using High Availability. The benefits are:
1. Not all the application load is hosted by a single server.
2. The secondary server is being utilized.
3. There is a very high probability the primary applications will execute on the secondary server.
4. The test and development applications are isolated from the production applications. If some catastrophe event occurs in development or training it does not impact production.
5. Maintenance and upgrades downtime do not impact production, test or development applications since the test and development applications can be temporarily hosted on the primary server.
Manually moving the test and development applications to the primary server tests the operation of HA and allows the system administration full access to the secondary server with disruptions. Once the changes have been made to the secondary server the test and development applications are moved back to validate full operation with the new changes. Once the changes have confirmed on the secondary server the production applications can be moved to the secondary server while the changes are performed on the primary server.
There are two downsides:
1. Management of the period support and maintenance scripts must be properly managed after a failover.
2. There may be additional licensing fees involved from the application(s) vendor(s). Please discuss the license issues with the vendors before making a final decision.
Conclusion
While it is more complex to configure, the Active-Active High Availability model is far more flexible for support and maintenance, utilizes all servers, and provides a much higher probability of no issues occurring after a failover. While there may be additional licensing fees, the uptime benefits of Active-Active normally outweigh the additional costs.
Santa Rosa Consulting is a national leader in providing integration solutions to the healthcare industry. We have the industry experts available to assist you in all of your integration requirements. To learn more about Santa Rosa’s expertise, please visit our website at www.Santarosaconsulting.com.