T 2.151 Lack of manufacturer support of applications regarding the use of virtual IT systems
The influence of the virtualisation technology on the design of computer centres outside of the mainframe world (IBM Z-Series, Siemens BS2000, SUN Enterprise 25000) only started to increase a few years ago and productive IT systems are virtualised to an increasing extent only since about 2005. Before, virtual IT systems were predominantly used in development and test environments. There is a large number of different virtualisation products additionally based on different technical approaches (server and operating system virtualisation). Therefore, no virtual IT system has been standardised up to this point in time in the way this is possible for IT systems based on x86 or x64 hardware.
Normally, applications are approved for a certain combination of operating system and hardware platform by their manufacturers. This means that they support the user of an application in troubleshooting, for example, if the application is operated with the corresponding operating system on the approved hardware platform. However, since the hardware platform "virtual IT system" has not yet been standardised, the application manufacturers are not able to make any general statements as to the extent to which the installation of their application on any virtual IT system is supported.
Virtual IT systems may be operated on the basis of completely different virtualisation technologies (server or operating system virtualisation) and therefore may be characterised by greatly differing properties. In virtual IT systems based on operating system virtualisation (SUN Solaris Zones, Parallels Virtuozzo), i.e. that may represent multiple instances of a single operating system, different, quasi operating system software libraries or different operating system cores cannot be used at all or only to a very limited extent, for example. Such a limitation normally cannot be found in virtual systems based on complete server virtualisation (e.g. Citrix XenServer, Microsoft HyperV, QEMU, Sun VirtualBox, VMware ESX) so that a generally applicable statement for all conceivable virtual IT systems regardless of which virtualisation technology cannot be made.
For the aforementioned reasons, manufacturers do not generally approve their applications to be operated on virtual IT systems, but only grant these approvals for certain combinations of operating system and specific virtualisation products if applicable. If it is not checked whether such an approval is present, there is the risk that support is rejected or restricted in the event of difficulties.
Example:
A large-scale company operates a comprehensive ERP system consisting of numerous servers. The ERP system consists of several database systems and approximately 30 application and 80 web servers. The company concluded a maintenance and support agreement with the manufacturer of the ERP system containing the promise of the manufacturer to provide support for occurring problems. The support agreement is subject to the condition that the ERP systems must be operated with the Windows Server 2003 operating system on typical hardware. For virtual systems, the manufacturer reserves a case-by-case review and does not issue any general approval.
The company wants to replace the server systems the ERP is operated on with new systems, since the existing systems have gotten relatively old in the meantime and hardware failures occur frequently. The administrators responsible report that the individual servers, especially the application and web servers, are not utilised significantly and that load peaks do not occur simultaneously on all systems, but are distributed to the systems over the course of a day. For these reasons, the decision to virtualise the application and web servers and to operate these in a virtual infrastructure consisting of several virtualisation servers is made. For the application servers, the company selects a server virtualisation solution, and for the web servers, it selects a product based on operating system virtualisation. Operating system virtualisation is deemed specifically appropriate for providing a large number of web servers, since major consolidation effects can be achieved in so doing, i.e. a very high number of virtual instances can be operated on one virtualisation server. For server virtualisation, the administrators do not expect any problems that may be attributed to the virtualisation software and assume that no virtualisation-related failures will occur.
After the ERP systems were virtualised without even consulting the manufacturer of the ERP systems, the ERP application runs without any failures for a certain time. After a couple of months, an employee recognises errors occurring in the storage module of the ERP software, however. It is determined that the incoming and outgoing goods entered by the warehouse employees using a web server were processed incorrectly in the warehouse. The ERP system now automatically triggers purchase orders although there still are sufficient goods in the warehouse. For other goods urgently required for production, the ERP system shows excessive stock levels, resulting in no replenishment orders being triggered and the production standing still. The company suffers damages to a considerable amount due to the loss of production.
The administrators of the ERP system address the problem and assume that there is a transmission issue in the interaction between web server and application server resulting in the incorrect processing. However, they are not able to find any solution and they consult the manufacturer. The manufacturer asks for the configuration of the ERP servers and automatically generated reports and reviews these in order to be able to narrow down and eliminate the error.
After the manufacturer of the ERP systems determined that the servers are operated on virtual platforms, the manufacturer informs the company that the ERP system runs on a platform that was not approved and thus not supported. The manufacturer determined a timing problem as the cause and refuses any further processing, since it is expected that the problem is related to the virtualisation of the systems. The manufacturer requests the company to reconstruct the operational scenario on non-virtualised hardware in order to rule out virtualisation as a cause of the problem.
The company is now forced to rent a large number of physical servers for reconstruction of the operational scenario. This reconstruction is very complex and time-consuming. Moreover, this significantly delays troubleshooting.
It turns out that the error also occurs on the physical systems and that it can therefore be largely ruled out that the server virtualisation was the cause of the error. Subsequently, the ERP manufacturer continues its efforts and the problem is resolved completely upon in-depth analysis.
The company using the ERP system now demands compensation from the software manufacturer for the costs incurred as a result of reproducing the error on physical servers, as well as the lost working time and the loss of production during that time that occurred while designing the parallel environment. According to the company, the solution of the problem was delayed unnecessarily by the software manufacturer, since it turned out that the virtualisation was not the cause of the problem. The manufacturer in turn refers to the wording of the maintenance and support agreement and denies any liability. Moreover, the manufacturer emphasises that the timing problem that lead to the assumption that the problem that occurred was connected to the virtualisation was only determined as a gesture of goodwill. The manufacturer could also have completely refused to processing the problem. The manufacturer of the ERP software wins the subsequent court proceedings.