It is amazing how much data gets captured as we interact with people and things throughout the day. Do you ever look at it?
I DO! I love data. I find it fascinating to see what trends you can observe.
The other day, I was reviewing how much water we use at home. Our city tracks usage, which I can view. For the past 12 years, my water usage was 48, 59, 46, 59, 56, 55, 62, 71, 62, 69, 73, 83.
Not super helpful until we turn that data into a graph.
Two things immediately pop out:
- Over the past 12 years, our water usage has steadily gone up, which makes sense as my kids get older and take longer showers (I can’t even imagine what happens to this graph when they become teenagers).
- Three points in time, my yearly water usage dropped. Why? After a little research, I concluded that each one of those years corresponds to the year where I replaced one of our 3 toilets, which were 20+ years old. You’ve probably heard that new toilets use less water than older ones, so much so that I can see the impact on my yearly water usage. Fascinating
For XenApp and XenDesktop admins, many of you know that Director includes real-time tracking and historical usage trends. Ever wondered how you can use this capability effectively?
Let’s say you have to install a cumulative update, a security patch or an app update on a XenApp host. Are you concerned about the potential impact on the scalability of the server? You should be. Who knows what it will do to the overall user density.
Because we are unsure about the potential impact of an update, we should follow a cloud-thinking strategy with updates by using a canary model where we patch a small subset of systems first to identify any issues before rolling out to all of production servers.
Let those servers run and bake for 7 days, if possible, to gather enough real-user data to generate useful historical trends in Director.
Here I did 2 hour historical trend, and although I can see increases in RAM and CPU, the insights can get lost with all of the minor fluctuations within real user behavior.
By using a longer time period for the trend, those minor fluctuations smooth out and better insights can be gathered.
Those insights are critical to the ongoing stability of your environment. Did CPU increase/decrease? Did RAM? Did user load? If there is an increase, you need to determine if the percent increase is greater than your available extra capacity within your environment. If so, you have to allocate more resources BEFORE you roll out the update to all production servers, or else you will end up with a lack of server resources to service all of your user requests.
In the end, if you don’t plan properly, your usability gets flushed down the drain.