In today’s “real time” world, people are getting used to accessing data instantly and can be easily frustrated without immediate access.
In some cases, it is not users’ frustrations we are dealing with, but a loss of profit when a system goes down due to a computer or human error. IT executives realize this, and seldom do they build their critical production systems without some type of high availability. How can you maximize your assets to provide the best available system?
In many cases, those assets could be a multi-building facility. In spite of what some challenges might bring by interconnecting different buildings, the benefits are substantial. Chances are you have some remote network closet on the campus that connects to your main data center, and yes, you could put some hardware in there. This is where IBM PowerHA stretched clusters come in. IBM PowerHA stretched clusters provides the same type of high availability that this solution has been delivering, over the years for AIX, but with participating nodes at different locations. This gives you the peace of mind that if one location goes down, the other one takes over. Did I also say that this was with a zero RPO (recovery point objective)?
We can achieve zero RPO because PowerHA stretched clusters work with AIX LVM mirroring, and the most common deployment model is a synchronous configuration (in Standard Edition). AIX is the one doing the mirroring at the LVM level, not PowerHA. PowerHA is there to monitor the systems, network, disks, and applications, among other things, and to be in charge of failover/failback. AIX can achieve mirror consistency between sites via what is called mirror pools. With mirror pools, you can guarantee that specific physical disks are participating in a volume mirror pair. When PowerHA fails over, all data will be “synched.”
You may be asking yourself what kind of distance is supported between our HA nodes, and the answer to that is, it depends. Remember, you need storage at your primary site and some secondary storage (probably a slower, less expensive disk) at your secondary location. Both nodes need to see both storage systems so they can achieve the cross-site mirror. If you have a direct SAN link from the local to the remote fabric, you may get up to 10-15kms. If you use some type of DWDM or SAN extender, you may get up to 300kms. Of course, you have to watch for application latency, since this is a synchronous configuration, and all writes have to go through the link before the I/O is complete.
So perhaps it’s time to make better use of some of those network closets in a different building inside your campus. When PowerHA is coupled with IBM PowerVM virtualization, you can have a single server running different PowerHA failover targets, achieving greater consolidation and providing cost savings. In the latest version—PowerHA SystemMirror 7.1.2—you can create stretched-clusters configurations with the Standard Edition. If you are thinking of having greater distances or asynchronous replication, Enterprise Edition is the way to go.
If you have more questions, feel free to reach out to me at firstname.lastname@example.org