Getting Started

 

If you are reading this blog, chances are that you consider setting up some form of monitoring for a web site or an information system that you are responsible for. You may have read about what types of monitoring there are, and you may be wondering where to start.


This article discusses a route that you could take. Of course, there are many other ways to do this. For this text I assume that you still have to get management support and some funding. For the sake of argument, I assume that you have no monitoring of any kind, and that your manager won’t simply hand you a wad of cash. I will show you how I would go about not only setting up monitoring for a system, but how to start a cycle of continuous improvement on your monitors and hence your insight in the machinery.




At a glance, the cycle is simple. Set up some reporting to get a little funding. Spend your money wisely to start a cycle of improving your monitors and expanding your capability to report interesting things to management and business.


Getting Started

The first step is to get your management to support and fund your efforts to improve the quality of the system. Your goal in this step is only to get your manager to talk to you about the monitoring of the system. The most effective way to do this is to set up some rudimentary functional monitoring and report some measurement results. Functional monitoring works well for management reporting, because it reports on functions of the system that non-technical people can relate to.


If the system is a web site that is exposed to the Internet, use one of the many cheap or free on-line web site monitors. For intranet applications or other systems, use a test robot or write a little script that performs a few operations and reports success and response times. Select an operation that you know is important to the users of the system. If you want to draw some extra attention, pick a function that you know has a tendency to stop working every once in a while, or is slow to use. Schedule the test periodically and collect the measurement results in a file, a spreadsheet or in a simple database.


From this data, you can start generating a monthly or weekly report. Send this report to a few key decision makers, accompanied by a short note explaining what the diagram shows. In the report, stay absolutely factual. Don’t attempt to interpret the results, draw conclusions or suggest changes. Curb the engineer in you. Report only the facts and how you got them. Let others come to you for more information, not the other way around.


Of course, you can’t just start sending out management reports to random managers. Print out the first version of your report and deliver it personally to your manager. Explain that you’ve set up a little measurement to see how that particular function of the system is doing. Tell him or her that you think this may be something that may be of value to the department and ask for permission to send a periodic update.


For those of you thinking they don’t need no stinkin’ permission to send out a report and for those thinking they don’t need to deliver the first report by hand: think again. This is headology.


Setting up this initial monitor and a template for the report should take no more than a day or so. If you find yourself spending more time than that, you may have lost sight of the goal of this first step: to open a dialogue with your manager. All you want to do at this stage is pique their interest. The rest will follow.


Talking Monitors

You will know that you are doing well when you start seeing your own graphs in other people’s documents and find your reports being forwarded to other people. Keep an eye on who gets our report and start planning for the second step: taking the first stab at serious technical monitoring.


Two things may happen once people start taking serious interest; you may be asked to report on more functions of the system or, if you have chosen to report on a brittle function of your system, you may be told to fix the problem. Either way, this means that you now have an opening to ask for resources to work with.


Split the man hours and money into three parts.

  1. 1.A part to automate and expand the functional monitoring. Functional monitoring is your antenna for problems. It triggers you when there is a problem in our system.

  2. 2.A part to set up some initial technical monitoring. You will need technical monitoring to track down and fix any problems that your functional monitor has exposed.

  3. 3.A part to make improvements to the system. You must show results if you want to be taken seriously next time you request a budget.


Be realistic when asking for resources. You can’t expect to get the money and man hours to set up a know-all-see-all monitor at this stage. Instead, try to get enough man hours to make some demonstrable progress on all of the three items outlined above.


It is my preference to use Open Source tools for monitoring. They are cheap and get the job done. I find that commercial products can only do more than Open Source tools when they are being operated by experts. It is hard to justify expensive tools and expert training based on a skimpy graph in an e-mail.


Rinse, Repeat

From here, repeat the exercise. Listen to the business and to management to learn what they find important functions of the system. Report on the performance and availability of those functions and use the reports as a lever to pry some more precious man hours from the department’s budget.

 
photo: Sanja Gjenerohttp://www.sxc.hu/profile/lusi
if you find this interesting, terrible or just would like to know more, e-mail memailto:kjkoster@kjkoster.org?subject=
web statistics
http://java-monitor.com/