An agro-processing organization with primary offices in Ballard, Washington and Winnipeg, Canada was in need of highly available systems with near 24X7 access. The business requirements called for redundant sites with optimal performance for the corporate office in Ballard and finance team in Winnipeg. Additionally, PLC (plan processing control software) systems were distributed into remote processing plants in eastern Washington and other locales.
Their primary systems included Microsoft Great Plains, Exchange Server, SQL Server and a Citrix farm. Two prior vendors had worked on the existing systems. The original vendor had setup the environment in a simple two-server failover Hyper-V cluster. Utilizing items such as cross-over cables, the vendor had not configured the system for growth. Additionally, they implemented end-of-life HP SAN equipment that did not support the installed version of Windows Server.
A second vendor tried to expand the cluster by adding two additional severs, but had not realized the limiting factors implemented by the original vendor. Additionally, their configuration of the new servers was not compatible with the original servers. Items such as the Hyper-V network name were not the same, disallowing features such as Live-Migration.
The final solution design called for a multi-stage approach. A new SAN & Cluster environment was built to accommodate the multi-site failover and tested comprehensively. After testing sign-off all data was then migrated.
- In each data center a new six-server Hyper-V cluster was built using Dell EqualLogic SAN systems, which were compatible with the current version of Windows Server and had hardware level replication capabilities.
- The six servers were split so that four were dedicated to production VMs and two for test. This was done logically, so that the test servers could be repurposed to production to handle increased month- and year-end processing loads.
- LUNs/disk partitions were created, some of which were intended for SAN-to-SAN replication. The non-replicated LUNs housed either virtual machines that had replication built-in (e.g. Exchange DAG or SQL replication) or virtual machine snapshots and smart paging files.
- New SQL Server Always-On installations were implemented to accommodate multi-site failover for Great Plains production and test environments, and the Citrix Farm.
- Data connectors were implemented to import plant processing data (e.g. tons of beans delivered and payments made) from the PLC systems into the Great Plains financial system
- Exchange mailbox servers were configured in a DAG (database availability group) to also accommodate the multi-site failover. Front-end Exchange servers were implemented in each location as well.
The final solution provided the client with the following benefits:
- System availability of 99.999%
- Real-time replication of data
- Fail-over from one data center to another with a complete copy of all data at the time of the failure
- Live migration of Virtual Machines across the cluster without loss of client connectivity
- Allowing server loads to be reallocated where necessary
- Allowing maintenance to be performed at any time of day