The new old cluster
We have been working for many years with the Protection business-critical applications through clusters. Installation and sale of complex Systems is certainly the first stage of the life cycle.
At this point, operation begins. A decisive parameter here is the Availability the Application.
Well, we didn't discover that, but our colleagues already recognized it in the great age of the Unix-operating systems.
The legendary VMS Cluster, Veritas Cluster Server for Solaris or the True64 Cluster have set high standards in this area. Availability set.
LifeKeeper was also available as an in-house product for NCR's MP-RAS and was one of the first solutions to be ported to the x86 world.
Microsoft Windows NT 3.5 became the preferred operating system basis - was that a coincidence?
Troubleshooting
As is well known, the availability cluster consists of at least two Serversto increase redundancy.
With a well thought-out architecture, the resulting overall system should no longer contain a single point of failure (SPoF).
The second and almost more important starting point for avoiding system failures is application monitoring.
And this is where the products sometimes differ greatly.
The more precise and flexible the monitoring of the Application can be adapted to the respective application, the more Error can be recognized.
The The best testing depth is usually achieved with solutions that Application be supplied by the customer.
However, if different cluster solutions are Application the right way for data center operation and can all specific adjustments be made?
There is no doubt that the overall failure of a cluster node brings clear conditions - but it is usually not so trivial. And without regular functionality checks, the expected result is not necessarily achievable in the event of an error.
Besides Hardware and the Application of course, the Data be protected.
Virtualization as a solution
Since there are a number of points to consider when setting up and operating a cluster in order to achieve the desired goal, server virtualization with its possibilities seemed to be a welcome solution, which would also reduce theAvailability the Systems with increased.
I would not dispute this, even though an additional virtualization layer has been introduced - in practice, however, this has proven to be stable.
The Possibility of virtualized Systems simply moving from one hardware node to another is undoubtedly a wonderful thing.
But how do I know when I need to do this, and will replacing the hardware solve the problem?
The Virtualization software detects both the failure of a Host-system, but can also be the operating system of the guest system monitor.
The Application remains outside
This is where our experience in application monitoring can be put to good use. It is certainly possible to implement a classic cluster as a virtualized system, provided certain conditions are met.
However, the option of relying on experience with the Protection of applications, and the switching of the Virtualization to leave.
In addition to the long-awaited simplifications, server virtualization has also (unintentionally) led to an enormous increase in the number of Systems can lead to.
Because the Application cluster, increases the number of Systems even further.
The idea of the One Node Cluster was developed out of this need.
Is it enough for many Errorthe system or even just the Application at monitor and restart it if necessary?
It certainly makes sense, but a nice name would have been part of it. One Node Cluster is like a single family - but we know in the IT already know what we're talking about, at least most of the time.
Maintaining an overview
To avoid introducing additional complexity and confusion into system administration, we pay attention to integrated Administration application monitoring.
It cannot be the goal of the Administration the Virtualization of the usually very many Systems to think about which Systems operated in the cluster or assigned to the various Host-systems and so on.
I already had this experience with one of our first cluster installations for the Protection a Database made.
The Customer wanted to use an offlineFuse his Database and held the Database on the first node - offline still didn't work: right.
Protection
With the ever-increasing dependence of almost all processes on the IT the willingness to deal with the disaster is growing - slowly.
The not quite cheap perfection certainly consists in a Backup–Data center or computer room.
Solutions are often offered that at least Data replicate to a second location. Only: The Data alone is just ones and zeros.
The next step would therefore be to also have the applications ready in the event of a disaster.
In addition to the possibilities offered by virtualization solutions, the cluster across data center boundaries - the so-called stretched cluster - is a good option.
It also makes sense to configure applications on physical Servers in the event of a fault with virtualized systems.
With the right framework conditions, there is nothing to be said against the virtual Data center in the Cloudto protect against disaster.
None of these solutions are as simple as the brochure suggests, as things usually turn out differently than expected. The technical basis is certainly the replication of the Data into the Cloud.
After the Installation the Application is also the access of users to the Cloud–Systems the next challenge. A multiple Test is an important success factor here.
SAP–Systems
We have not even considered the SAP–Systems. As highly business-critical Systems they are naturally predestined for application protection.
There is also enough SPoF on the individual system.
As a commitment to application availability, the High AvailabilityCertification which is supported by the LinuxLab of the SAP was introduced last year.
Up to this point, cluster solutions were classified as middleware and were not assigned to any Test subjected to.
But not only the technical test catalog is the subject of the SAP–Certification for cluster products.
For us and our Customers Particularly important was the Certification In the event of a fault, the associated commitment to continue to operate in this environment without deactivating the cluster components Support to perform.
To this end, a defined support process, the definition of which is part of the Certification is to integrate the providers of the availability solution into the support process.
Another important component of cluster certification is the possibility of SAP-clusters via SAP Adaptive Computing Controller (ACC) or Landscape Virtualization Management (LVM) can be controlled.
Together with our colleagues in St. Leon-Rot, we are currently investigating whether it makes sense and is practicable, SAP–Systems with a Cloud-solution to protect against disaster.
Conclusion
The emperor is dead - long live the emperor.
No, I don't want to claim that clusters will continue to exist forever in the same way they were developed years ago.
In addition to its importance for disaster protection, knowledge of application monitoring is increasingly becoming the focus for increasing the Availability.
However, it is also important that the availability solution can be integrated into management systems that Hardware or virtualization environments monitor and control. If you keep this in mind and try to master the complexity of the overall solution, clusters continue to be a good means of Fuse the availability of applications.
The modern emperor lives.
