Information Technology Reference
In-Depth Information
a planned outage or upgrade. You inhibit alerts to conditionally cause one alert to not fire
when another condition is active.
17.5 Visualization
Visualization is the creation of a visual representation of one or more metrics. It is more
than just creating the pretty graphs that will impress management. Visualization helps find
meaning in large amounts of data.
Awell-designedmonitoringsystemdoesnotcollectdatatwice,onceforreal-time alert-
ing and yet again for visualization. If you collect it for visualization, be able to alert on it.
If you alert on it, store it so that you can visualize it.
Simple graphs can display raw data, summarized data, or a comparison of two more
metrics. Visualization also involves synthesizing new metrics from others. For example, a
rate can be calculated as the function of a counter and time.
Raw data graphed over time, either as a single point or as an aggregate of similar met-
rics, is useful for metrics that change slowly. For example, if the memory use of a server
over time, which usually remains nearly constant, changes after a particular software re-
lease, it is something to investigate. If the usage went up and developers did not expect it
to, there may be a memory leak. If usage went down and developers did not expect it to,
there may be a bug.
Countersarebestviewedasrates.Acounter,visualizedinitsrawform,lookslikealine
going up and to the right. If the rate over time is calculated, we will see either acceleration,
deceleration, or an equilibrium.
Some visualizations are less useful and, in fact, can be misleading. A pie chart of disk
space used and unused is a common dashboard fixture that we find rather pretty but point-
less. Knowing that a disk has 30 percent space remaining isn't useful without knowing the
size of the disk. Seeing two such pie charts, side by side, of two different-sized disks, is
rarely a useful comparison.
In contrast, calculating the rate at which the disk is filling and displaying the derivative
of that rate becomes an actionable metric. It helps predict when the disk space will be ex-
hausted. Graph this derivative against the derivative of other resource consumption rates,
such as the number of new accounts created, and you can see how resources are consumed
in proportion to each other.
The only thing we dislike more than pie charts are averages, or the mathematical mean.
Averages can be misleading in ways that encourage you to make bad decisions. If half of
your customers like hot tea and the other half like cold tea, none of them will be happy if
you serve them all lukewarm tea even though you have served the same average temperat-
Search WWH ::
Custom Search