Monitoring and managing server performance isn’t just a matter of gathering metrics. A server’s performance can...
directly influence workload performance. This affects business criteria such as user satisfaction, service-level agreement obligations and regulatory compliance.
Server performance is monitored through software tools, but selecting the best server performance monitoring software for a business can be a complex and time-consuming process when business results are on the line. There are many server performance monitoring software products to choose from, and they span an enormous range of cost, complexity
Features and capabilities of performance monitoring software
One of the most difficult hurdles to overcome when selecting server performance monitoring software is deciding what to monitor in the first place. Modern server performance monitoring tools go beyond the CPU, memory and I/O metrics gathering of legacy tools, such as Microsoft's PerfMon. Establishing the scope of the tool upfront will make it easier to weed out products that are too simple or too complex for your monitoring requirements.
Consider support for important foundation technologies such as virtualization. For example, the tool should support the prevailing hypervisors such as Hyper-V and ESXi, and it should be able to report on the health and performance of both physical and virtual servers. Many tools offer sophisticated capabilities that cover application-specific monitoring as well as monitoring for the network, storage
A platform such as SolarWinds Worldwide LLC's Server and Application Monitor can abstract fundamental hardware metrics in favor of application-level monitoring. It can map infrastructure dependencies to help with tasks such as capacity planning and root cause troubleshooting, and it can provide detailed monitoring of more than 200 enterprise applications or application platforms. Server and Application Monitor can also track server storage volumes, disk usage
Heterogeneous or homogeneous
Consider the hardware environment that the tool will support. A data center built around a homogeneous hardware environment often uses the vendor's suggested management and monitoring tools. For example, hyper-converged infrastructure (HCI) platforms usually use the HCI vendor's monitoring and management tools to great effect, whereas a more traditional heterogeneous data center might make better use of general-purpose server performance monitoring software.
For example, the Cisco HyperFlex family of HCI platforms delivers its best management results using the Cisco HyperFlex Connect platform, and Nutanix platform users would likely opt for the native Prism tool for HCI management. In both cases, the software tool is tailored to the hardware platform, which results in more convenient and accurate reporting.
A more traditional data center built with a mix of vendors' products and platforms generally has more monitoring and management software options; it’s important to evaluate the accuracy and granularity of the metrics and other data that the tool collects.
A tool that must accommodate more heterogeneous systems is more likely to overlook, omit or misreport more system details. This can force system administrators to use multiple tools to accommodate any incompatibilities or integration shortfalls. Ideally, administrators should seek a single pane of glass for systems management tasks, such as server performance monitoring. Multiple tools that do the same job on different systems are not desirable.
Data collection mechanisms
Evaluate the data collection scheme that each server performance monitoring software uses, and consider the implications of those schemes on everyday operations. A tool should be able to collect data from each system natively without supplemental data gathering mechanisms. This agentless operation is commonplace for homogeneous environments where tools are tailored to the underlying systems. The tool is written to gather specific data from a limited range of hardware and software, so it doesn't need additional mechanisms. Unfortunately, such tools cannot readily support systems and software that falls outside of that intended scope.
Software agents support a wider range of systems and software. Agents are small software applications that are tailored to run on certain systems, collect desired data such as a performance metrics and dependencies, and report that data back to a central repository -- such as a monitoring server -- which organizes, processes and visualizes that data. Agents are extremely common for monitoring heterogeneous environments, and they can be tailored to almost any system. Agents provide versatility, but they also present additional workloads that IT administrators must manage and maintain. Bugs can occur, and admins must periodically apply patches and updates. This adds to the workload management tasks that system administrators must handle on a regular basis.
Plug-ins or management packs, which install within the tool itself and extend its capabilities with additional systems, are an alternative to agents. Plug-ins are often used to support particular applications or environments.
For example, Microsoft System Center Operations Manager (SCOM) uses a plug-in to extend to manage resources in the AWS public cloud. Similarly, SCOM can use a management pack to support Cisco Unified Computing System platforms. ManageEngine OpManager offers an application performance management plug-in capable of monitoring application availability, health
Another approach to system performance data collection is using APIs. A system or software platform collects certain data and exposes it through predefined APIs, and developers create software that includes an SDK or
Consequently, APIs are most often used to integrate multiple software tools rather than directly accessing and collecting data from servers or other hardware. Still, admins must consider API versions and compatibility when selecting and evaluating any tools that rely on APIs.
Storage and networking
Server performance monitoring software can produce a substantial volume of data in the form of metrics and logs. This can put significant demands on networks and storage.
For example, the traffic produced by hundreds of servers that deliver high-resolution metrics in real time can potentially strain limited network bandwidth. It's important to consider agents' bandwidth needs -- especially as agents proliferate across virtualized systems -- and evaluate the best means of moving collected data around the environment. For example, some tools might collect data locally and only periodically move bulk data across the network to a central server for processing and reporting.
In addition, raw metrics and logs must be stored, often in a supporting database such as SQL. Stored data must then be processed, visualized, and ultimately retained and protected in accordance with prevailing regulatory policy. All of this storage and retention demands ample disk capacity and attention to storage subsystem performance.
For example, a server performance monitoring platform might need a RAID volume to meet storage I/O needs. Large server environments might demand an additional investment in storage capacity and data protection to support a comprehensive server performance monitoring deployment.
The more tools that admins must handle, the more time-consuming and error-prone each task becomes. Although single pane of glass management remains an elusive goal for many organizations, tool integrations are systematically improving through mechanisms such as APIs and plug-ins.
These integrations allow tools to share data and reporting to create a more ubiquitous management environment. Admins must consider how the server performance monitoring software they choose interoperates with other monitoring and management tools that might be present in their organization.
Integration is rarely native or automatic, and admins should expect tool integration to involve some level of work to resolve sometimes-subtle differences in data formats or communication preferences. For example, administrators might prefer to integrate two tools through simple network management protocol
Many organizations embrace public and hybrid clouds, but public cloud providers expose infrastructure details -- such as resource provisioning, usage
For example, SolarWinds Server and Application Monitor can monitor local infrastructure along with infrastructure in AWS and Microsoft Azure. Even if the business plans to utilize public clouds at some point in the future, it might be worth future-proofing the new tool by selecting a monitoring tool that includes cloud support.
The responsiveness and quality of a vendor's support are often overlooked -- until something goes wrong with a product. But when an organization's infrastructure relies on a monitoring tool, it's usually worth investigating and comparing that vendor's support options.
Evaluate telephone, email and web-based support. Check for training materials and guides. Review release notes and product roadmaps. A world-class tool is useless if it's not working.
Server performance monitoring software must be able to deliver the information that administrators need in a useful manner. What that looks like is different for every business, so it's important to consider the customizations that each prospective tool provides. The ability to toggle desired data points on or off; change the location or emphasis of each data point, such as making important KPI graphs larger than less-important ones; or select alternative display styles -- bar graphs versus pie chart, for example -- can make a huge difference in the readability of relevant performance information.
For example, Zabbix LLC provides extensive customization of visualization elements, such as display layouts, graphs, network maps and even slideshows that switch automatically between various configured screens.
IT deploys traditional server performance monitoring software on premises, but an increasing number of tools are available as cloud-based services. For example, the Cisco AppDynamics platform is available for either
On-premises deployments offer organizations more control and ownership, but the company also bears all of the responsibility and costs for the tool and its operation. Adopting the tool as a service imposes a monthly cost that may vary based on the size of the infrastructure it monitors, but the organization sheds the resources and maintenance needed to support the tool. It's important to consider the pros and cons of each deployment option when evaluating a prospective tool.
Ultimately, the selection of server performance monitoring software comes with a price tag. Commercial software costs often include an upfront licensing expense, followed by additional costs for support and updates. But savvy buyers can potentially strike attractive deals with eager vendors seeking to gain or expand a presence in the enterprise. Don't be shy about negotiating for better licensing and support terms.
There are also popular open