Server performance monitoring tools can simplify systems management tasks by allowing administrators to see the entire environment through a single interface and ensure that each real or virtual workload is performing at an adequate level. But selecting the best tool for your unique needs and budget can be an enormous challenge. The following questions summarize the most common issues leading up to a new tool acquisition.
What server performance monitoring features should I focus on when selecting a product?
When evaluating server performance monitoring products, the ultimate goal is to select one that meets your business needs and budget. Many tools share some common features, such as CPU or
One of the key reasons server monitoring tools are frequently underused or abandoned is because businesses choose the wrong tool or choose one that is simply too sophisticated and costly to manage with available IT staff.
Start the search by selecting a tool that supports your current server hardware and future upgrades. For example, if your company uses Dell PowerEdge servers -- and plans to continue -- be sure that the monitoring tool is suited for those servers and verify that the vendor will continue that support into the future. Hardware compatibility helps ensure that the tool can monitor hardware-level behaviors, such as temperatures and fan speeds, and allows staff to locate potential problems before actual system failures occur.
The standard feature set usually includes such basics as CPU, memory, disk, network I/O utilization and process monitoring. Network characteristics are usually included, allowing staff to monitor each machine in the corporate LAN, evaluate the amount of network load handled by each server and keep a close watch on TCP/IP services. An increasing number of monitoring tools provide support for servers with VMware, Microsoft and Citrix virtualization, automatically detecting virtual machines and the resources provisioned to them.
Beyond the basics, look for tools with alerting, reporting, inventory and capacity-planning abilities. Inventory abilities allow the tool to inspect and report on each system's hardware and software configuration of such common applications as application servers, database servers, Web servers, middleware and so on. This can be important when planning for hardware upgrades and software license renewals. Capacity planning watches server resource use over time and helps to predict resource requirements into the future. This allows a business to budget for growth in computing needs before resource shortages cause provisioning problems.
What is the difference between agent and agentless monitoring products? Which is better?
Server performance monitoring is typically handled through a client-server computing architecture. For example, the monitoring tool and its associated software (such as an SQL database) are installed on one or more dedicated servers; the tool then discovers and communicates with other servers in the environment, prepares reports and delivers performance data to the user interface. There are basically two ways to discover and communicate with other servers.
The first is to install agent software on every server being managed by the monitoring tool. The agent runs as a utility on each server, gathering and preparing information that can easily be passed back to the monitoring server. Agents usually offer more flexibility, allowing a wider variety of servers to be managed. Unfortunately, agents require some computing overhead on each server, potentially slowing the performance of each managed server. And agent software must also be maintained and upgraded as needed, which adds to staff workload.
The alternative is to use agentless monitoring. This means the monitoring tool itself can communicate directly with the server to get status information and exert control over the system. This approach is easier to maintain because there are no agents, so there are no software version problems. But agentless monitoring depends on hardware-level awareness of each system, so the monitoring tool's usefulness is limited to the server makes, models or supported operating system versions. For example, SolarWinds Server and Application Monitor currently supports Hewlett-Packard ProLiant, Dell PowerEdge and IBM X Series servers.
Ultimately, a homogeneous shop running a limited suite of systems can probably get better results from agentless performance monitoring, while a heterogeneous organization may need an agent-based approach to accommodate the variety of systems running in the environment.
My job is on the line, so how can I tell if a server performance monitoring tool is the right choice for my business?
Acquiring even basic monitoring tools can involve a substantial amount of work and expense, and the investment is worthless if you can't monitor and control the systems that you need. This means there is absolutely no substitute for your own careful product evaluation and proof-of-concept work.
Vendors routinely provide free evaluation versions of their products. Before making a purchase commitment, get tools in-house for testing and evaluation. Go through the installation and setup. Make sure that you can monitor every system make and model that is running in production. Also, verify that the tool can handle every operating system version within your environment. Check the alerting and reporting features and so on.
Also, perform some due diligence on the vendor. For example, try their service and support, or see if they can help you customize the tool. Their responses can be a good indication of the kind of service that you might net after making a purchase.
Server performance monitoring is emerging as a critical task for IT administrators, but the choice of monitoring tool should not be made lightly. It's important to match the capabilities of each tool to your needs, environment, budget and available IT talent. And before making an actual purchase, take the time to test and evaluate potential candidates in a lab setting. This affords an opportunity to see how they work and gain valuable experience with each tool without impacting the production environment.
This was first published in February 2013