In the first installment of this series, the requirements and challenges tied to desktop inventory were described. In this blog post, the server side will be covered. The datacenter environment has the benefits of being smaller and more stable than the desktop. However, software license management in this environment requires additional data—from hardware properties to relationships between hosts, virtual machines, partitions and clusters. Some complex server license models require additional data. Software identification is also more challenging due to the heterogeneous nature of the datacenter and lack of standards.
In the desktop environment, only a few operating systems are found: primarily Windows, MAC OS and mobile device operating systems. In the datacenter, many more exist—from Windows Server to many variations of UNIX and Linux or Mainframe systems. As servers have a longer lifetime than desktops, it is not uncommon to find versions of operating systems that are more than 5 and sometimes up to 10 years old. Inventory agents from configuration management or inventory tools must be able to support all of these platforms. Identifying software products is more challenging too as there is no unique installation repository, such as the Windows Add/Remove Programs area, to register installations. Multiple installation technologies frequently coexist, such as Red Hat Package Manager (RPM), Oracle Universal Installer (OUI), Flexera Software InstallAnywhere, etc. Some vendors install products with no registration at all; in this case deep analysis of installed files is needed.
In the datacenter, credentials and access rights are often difficult to obtain. It is fairly easy, when working in a Windows desktop environment, to have access to domain admin rights, as operators performing the discovery and inventory are usually granted these privileges. In the datacenter, getting access to the different Operating Systems, databases or applications requires additional efforts. There is always a concern that an inventory agent may disrupt the server operations, either by using resources or for security reasons. Compared to the desktop world, additional documentation or explanations need to be provided to system administrators, and detailed impact analysis or validation tests performed to comply with the organization’s processes. In some circumstances, running an inventory on mission critical systems is not an option and only a manual inventory relying on information provided by system administrators is possible.
Many licensing metrics are tied to hardware properties. The hardware inventory data needed for servers includes:
- The processor type, the manufacturer, the model and configuration. For instance, you may need to know whether hyper threading is active or not.
- The number of processors
- The processor speed
- The number of cores per processor
- The number of sockets. This information is usually not available through inventory and may need to be manually entered into the asset management system
- There are a few corner case scenarios where additional data is needed, such as the amount of memory, disk size, etc.
Virtualization technologies such as VMware vCenter/vSphere, Microsoft Hyper-V, etc. coexist in the datacenter with hard partition technologies: vPar, nPar, LPAR, Solaris Zones, etc.; each of them dedicated to a server platform. License metrics, such as IBM Processor Value Unit (PVU) in a full capacity configuration or Oracle Processor, use hardware properties of the underlying host supporting the virtual machine and require data on the processor pool when hard partitions are used. Understanding the relationships between virtual machines/partitions and their host, and if applicable, the structure of the processor pools is required in these scenarios. In some cases, such as VMware, an inventory agent has no visibility on the host and a direct connection to the VMware application programming interface (API) is required.
Applications in the datacenter communicate with other servers or the outside world over the network. Port scanning in this environment can be used as a discovery mechanism to uncover many of the software products installed on servers; sending a query to the appropriate port returns an acknowledgement enabling the discovery of the application. There are ways to go further with many products as they often support some type of remote API or connection. In this last case, additional credentials are typically required. For instance, VMware ESX/ESXi servers can be found by querying port 443 on ESX/ESXi, VSphere or VCenter servers. Additional queries to the VMware API enable the discovery/inventory of VMware hosts and virtual machines along with the VMware configuration and components. Products can be running on a machine without installation and in this case, only usage tracking or network discovery will unveil the use of these products. An Oracle database instance can be configured to be run from a shared directory, for example.
In many cases, datacenter software products require additional inspection due to their specific licensing model. It may not be enough to discover only the instance of a product; measurements are required for the license compliance calculation. For instance, SAP applications use a mix of metrics based on named users and on capacity used by packages (engines). Only specific queries sent to SAP systems will return the data required for software license management and optimization. The same applies to Oracle databases licensed using the processor metric. In this case, specific queries measuring the use of management packs and options must be run against the database to accurately calculate license compliance.
Vendors, such as Amazon, Rackspace and Microsoft (Azure), enable customers to extend or move their datacenter to the cloud through Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) offerings. Many publishers, such as Microsoft, IBM and Oracle, enable their customers to move their existing on-premise licenses to the cloud, on servers belonging to and managed by third party vendors. In this case, all the licenses moved to the cloud must be accounted for. These vendors’ offerings may also include product licenses as part of a Platform as a Service (PaaS) service. For instance, as part of a subscription for an Amazon server, an Oracle database might be included. Thus, an organization may be in a situation where it is using both its own licenses and subscription based Oracle licenses in the cloud. The use of a combination of “traditional” datacenter discovery and inventory tools, as described above, and APIs from IaaS and PaaS vendors would enable an accurate inventory of the cloud based servers.
Datacenter inventory is very different from the desktop world. Despite a smaller number of devices, it is a more complex and challenging environment due to the diversity of operating systems and technologies, the licensing metrics requiring additional data, and the emergence of hybrid (public-private) and private clouds. Different techniques must be used; a device based inventory will likely fall short of accurately reporting inventory and usage data. As in the case of the desktop world, only a combination of tools will provide the data required for license management and Software License Optimization.
***
To learn more, please read our white paper: Optimized License Management for the Datacenter