Time to Rethink Your Monitoring and Log Analytics Strategy — 3 things to consider

Ben Newton
Feb 2, 2018 · 6 min read

How you ask? I think that the wisdom of my toddler can come to our rescue! Let us sit on his Disney character play-mat, and learn from him. Come. Sit down. Relax.

I think raising a Toddler makes doing pretty much anything else look easier (… and teaches me humility) — so why not start from there. I will do my best now to take some of the core wisdom my toddler teaches me every day and apply it to the relatively simpler exercise of outlining what a highly scalable analytics service should look like… Here we go!!!

Toddler Perspective #1: Veggies with Cheese on Top are still Veggies

As anyone with a small child knows, food is a battleground. There is no force on earth that can make my toddler eat what he doesn’t want to eat, but he can hear the tell-tale crackling of a potato-chips bag from 50 yards away through brick walls. In the same way, he is not fooled by our weak attempts at deceiving him as to what we put on his plate. “No, Daddy. No like it”. It does seem that, as we grow older and more “mature”, we lose some of that clear-headed embrace of the facts right on our metaphorical plates — if I can stretch the analogy a bit.

Within the context of the current topic at hand, we (with a big, community wide W) seem quite adept at convincing ourselves that we can dress up software meant to solve other problems, run against other data sets, and function in other context s— and just re-label it. So, let’s be honest with ourselves. Taking installed software and labelling it as “Something Something Cloud” does not a modern analytics solution make. In the same way, taking a log-analytics solution and relabeling it as a time series metrics solution, or vice versa, just doesn’t work. My toddler expects better effort from us, and you should expect it too. So, put the effort in to make Kale Chips, and don’t disappoint everybody, including yourselves, with melted velveeta cheese on broccoli. “No. Me no like it”.

Toddler Perspective #2: Words Matter

Before having kids I never appreciated the amazing process that learning language is. Watching my kids advance from “Goo goo gah gah” to full sentences was a trip. My youngest, the current toddler, has had more of a challenge because he has struggled with some speech impediments and delays. So, it is sometimes heartbreaking, but also sometimes down right hilarious, to watch him communicate. He desperately wants to be understood, and he just keeps working away at it. A perfect example happened yesterday during bathtime. My son was telling his older sister, in his command voice (think Darth Vader interspersed with dinosaur style roars), that he would “<inaudible>ck you”. I was more intrigued than angry and asked him to repeat himself. My ears were still hearing more Samuel L. Jackson-style vocabulary than Elmo at this point. Over dinner, we continued to explore this situation, and uncovered that my son was dropping the “s” and converting the “m” to an f in the word “smack”. Ok. That seems much less horrifying and significantly less like a damning indictment of my skills as a father.

So, the point of that story is that words matter — particularly who says them, and the context in which they say them. In the brave new world of custom instrumented, microservices-based architectures context is everything. You can’t depend on the tired old server-based models and architecture to guide your teams towards uncovering thorny issues and resolving complicated problems. Just like my toddler gets frustrated when people don’t understand what he is trying to say, analytics tools without a concept of your application context and metadata are hampered at best, counterproductive at worst. You need to be able to stamp the unique identity of your data onto that very same data, and then use those unique “signposts” to guide your team when there just isn’t time to figure things out the hard way.

Toddler Perspective #3: A remote controlled car without the remote isn’t much fun

If there one thing that Toddlers don’t excel at (other than basic sanitary concerns), it is keeping up with stuff. I don’t know how many times I have asked my son where some particular toy is, only to have him stare blankly at me and wander aimlessly for the next 5 minutes looking for it (we inevitably find it 3–6 months later in some random cubby hole somewhere). This is particularly concerning with his remote controlled cars. He got one cool one in particular for Christmas that has not one, not two, but three components! He had misplaced them all with an hour of opening the present. Now, we usually hunt for the pieces at least daily (I’ll admit that is partly because I like to play with it, but that’s besides the point. Focus!). Without the remote, this remote controlled police car is really just a annoying light show that startles you when you are visiting the restroom on a dark night (the Aliens have finally come for me!).

Now to continue that analogy, our industry has spent the last decade peddling the analytics platform version of remote controlled cars without remotes to control them, or even bits and pieces of a remote controlled car — assembly required. Integration is the name of the game now. And this broken system was really just a reflection of the ineffective organization of the development and operations teams themselves that used the analytics (Conway’s law anyone?). If we now stretch the analogy to its breaking point (woohoo!) — in the old days there was one team that took the car out of the toy box, one that placed in it on the floor, another that was in charge of batteries, and yet another that had the experts on the remote (and refused to let others touch it). The multi-week process of readying the car for fun was mapped out with beautiful Gantt charts developed over dozens of highly “productive” meetings. By contrast, in today’s world of modern apps, by the time the laptop running Microsoft Project has booted up, the integrated DevOps-style team has already complete two laps of the house with the remote controlled car. These new teams don’t tolerate the tired old silos. If you don’t get your collective &^%$ together, they will just get some open source thing up and running in an afternoon, and make the best of it. They don’t even know what a Gantt chart is.

So, in conclusion, my toddler wants everybody to be happy and productive (and slip him candy when Daddy and Mommy aren’t looking). He’s a straight shooter, and he doesn’t truck with weak attempts at meeting his imperial needs. He expects uninterrupted devotion, creativity, and strong effort when fulfilling his needs. And that’s what our users expect us, dear readers. This modern world of high-performance business just doesn’t have the time to muck around with Enterprise software — They “no like it”. Time to put your big boy/big girl pants, and get to work!

Newtonian Nuggets

Thoughts on what's going on in technology, data, analytics, culture and other nerdy topics

Ben Newton

Written by

Proud Father, Avid Reader, Musician, Host of the Masters of Data Podcast, Product Evangelist @Sumologic

Newtonian Nuggets

Thoughts on what's going on in technology, data, analytics, culture and other nerdy topics