When it comes to business, data is everything. Whether it’s sales, supply, marketing, or your I.T. systems, every day is a constant stream of decisions to be made. To make the right decisions, you need enough knowledge. To have enough knowledge, you need the right data.
What’s the best way to collect all this data? In the information age, there’s so much data out there, all the time. It can be hard to know what you should be looking for and what monitoring and alerting strategy to employ so that you’re getting data that’s actually useful.
In this guide to monitoring data, we’ll walk you through the five basic questions you need to ask—why, what, when, where, and how—that will drive real business insights, intelligent decision making, and rapid disaster response.
Why Monitor Your Data?
As mentioned above, data drives smart decision making. If you want to know how your network, or supply chain, or sales team is doing—and be able to act accordingly—you’ll need adequate information in the first place. Monitoring data means that you’re collecting data so it’s ready for a decision. It means that when you need to make a decision, you don’t have to scramble to find the right data. You already have it on hand!
In general, as you monitor, you’ll be collecting data for three types of decisions. These are:
- Performance – The efficiency and effectiveness of your current system, whatever it may be. With good data, you can identify bottlenecks, make sure you’re meeting targets, and make informed decisions on how best to upgrade the system and improve performance.
- Diagnostics – When something in your system goes wrong, you’ll need to rely on the data you’ve got to fix it quickly and at the lowest cost to you. By using a good monitoring and alerting strategy and getting the right data in a timely way, you can make the right decision on current and potential errors.
- Security – Internal and external security of your systems is more important than ever before. Analyzing patterns in data can find unusual activity, improper access, and record manipulation. It can also indicate the best ways to increase security and develop effective security responses.
What Type of Data Should You Monitor?
All data can be divided into two categories: qualitative and quantitative. You will need to collect both in order to get the full picture of your system.
Quantitative data is numerical data. It’s often also called “metrics” since it consists of numerical measurements. The main benefit of application performance metrics is that, because they’re numbers, it’s easy to compare data against other data or against a set goal. Although metrics may not tell you why something is succeeding or failing, analyzing the data will allow you to quickly see how well your systems are performing.
Examples of quantitative data—or key infrastructure monitoring metrics—for servers and networks include:
- Work data – Such as how much work is being done by the system over time, how quickly something (e.g., a data query) is completed, how often it succeeds, how often it fails, and so on.
- Resource data – Such as how long a resource is used for, how often it’s accessed, how often it’s requested but not available, how often it reports an error, and so on.
Qualitative data, on the other hand, is non-numerical data. It describes events, descriptive reports, or changes within a system, and it is most often delivered in log format or as an alert.
While quantitative data doesn’t necessarily tell you about the performance of your system (except in general terms of success/failure), it can provide excellent insight if something goes wrong, allowing you to diagnose errors, track failures, and be warned against future problems.
All qualitative data should be monitored, with logs stored for future analysis and data alerting prioritized for urgency, and possibly redirected to an appropriate employee.
When Should You Collect the Data?
One of the fundamental ideas behind data monitoring is that, as long as you have the resources, it should be done all the time. In many cases, you won’t know what data you need until you need it, and it’s much better to have to sort through too much data than to have not enough data for a decision.
Having said that, there are two important factors to consider that will impact your resources:
- Granularity – In other words, how often the data is collected. Recording application performance every hour, for example, isn’t useful, as use can vary significantly during that hour. On the other hand, recording application performance metrics every millisecond may itself put a heavy burden on the processor, slowing the system down and distorting the data. This is a balance that will have to be determined on a case-by-case basis, depending on the data being monitored.
- Length of time – When will you need to access or analyze the data? How long should you store the recorded data? In most cases the longer the better but, again, it depends on the type of data and available resources (like storage space). It’s normally recommended that you store data for at least a year, so you can perform comparisons that take annual/seasonal variation into account.
Where Should You Monitor Your Data?
The scope of your data collection is also an important consideration, especially if you manage systems in multiple locations. When it comes to metrics, good data should be easily comparable, either to benchmarks or to similar data from other sources. So, if you’re collecting data from multiple locations, you need to make sure you’re monitoring the same sources in each.
How Should You Monitor Your Data?
Data becomes important not when you collect it, but when you use it to make a decision. A critical part of your data monitoring, therefore, is to make sure your data is clean, proper, and ready for use as soon as possible before the data is actually needed. That way, when it becomes important, you’re not wasting time or money sorting through data to make it useful or readable.
In other words, your monitoring and collection techniques need to produce the right data. These techniques should include:
- Labeling – Stored data should be labeled comprehensively, including date of collection, type of data, location of source, and so on. This makes it easier to sort and much easier to find quickly later on.
- Categorization – Like labeling, categorizing your data makes it easier and faster to retrieve at a later date. It also makes it simpler to compare with like data. Use tags for categorization so that the data can belong to multiple scopes and datasets.
- Cleaning – Quantitative data needs to be able to be read by algorithms and other software so that it can be compared and compiled into reports without error. Quantitative data should also be readable by an average user, not just someone with technical knowledge of the system. If either quantitative or qualitative data doesn’t meet these criteria, it should be cleaned up before anyone attempts to use it.
- Automation – Data monitoring, collection, and storage should be automated as much as possible. This removes the possibility of human error (or deliberate actions) tainting the data and making it unusable. Although it may take some effort to set up, automation will save a lot of time in the long run.
System Monitoring Made Simple
With so much data available to today’s administrators, staying on top of your monitoring and alerting strategy can seem overwhelming. Power Admin makes active network monitoring rapid and intuitive with PA Server Monitor. Capable of capturing metrics across hundreds of local and remote sites, PA Server Monitor is a versatile, robust server monitoring service that puts the power of data back in your hands.