Analysis of Dmitry Stolyarov's report on monitoring Kubernetes

    Oh, more than a year has passed since the previous analysis, but the time has come to return the tradition. Today we look at the report of Dmitry distol Stolyarov “Monitoring and Kubernetes”, which was presented at the RIT ++ festival in May of this year, and we are trying to understand why people usually like Dmitry's reports.

    Slides (all 344 pieces) can be found here .
    Disclaimer : about monitoring and Kubernetes is not the article itself, but a report that understands it.



    Dmitry begins his story with three stories. Their meaning is to set a common understanding for the audience and speaker that good monitoring should be able to. This is an important point: it is necessary to clarify the requirements for the system, but most speakers do not try to give them an interesting and memorable form. The list of bulletins of the type " Timeliness of reporting of incidents " in this place is very common and leads to the loss of half of the audience. Stories are much better because they are more interesting .

    I really love the Congress Hall of SKOLKOVO Business School (the main hall of RIT and Highload ++) because it shows very clearly from above how many devices with glowing screens are included in the hall. This is a very simple and reliable metric on whether people are interested in watching the scene. In our video, the general plan of the hall happens a bit later than the collection of requirements, but it can be seen that the interest has still not faded. Consider the frame from 14:58:

    I am observing here one computer is turned on, and its owner is clearly looking at the speaker. This coincides with my impressions: the introductory (often boring) part of the report was presented in such a way that it was interesting to watch and did not want to be distracted. Stories played an important role in this.

    I note that in stories Dmitry actively uses metaphors, and I do not really like this technique. The fact is that the metaphor is not always and equally understood by everyone. In particular, in secondary mathematics, I dare to hope, I am able to do pretty well, and what happened to the car and the odometer is obvious to me. The connection between this and the topic of the report is equally obvious to me. But I have never been engaged in monitoring as such, and what exactly should be the analogue of a blood test and tomograph remained a mystery to me. I’m not a target audience, but with examples of what could be tracked, but they don’t do it now, even I would understand everything.

    In general, in my opinion, a metaphor is often a weed that is better to mercilessly weed out.


    In the video, the conclusions section starts at 49:28 . Dmitry recalls the requirements for a well-tuned monitoring system (this is important), and Kubernetes innovations (this is also important). But not the fact that such reminders work well. Still, on both of these slides, six or seven points, we did not repeat them so often during the report so that they would become entrenched in the memory, and in the end, the audience does not have time to comprehend them again. It seems to me that the goal “to make the audience remember how to relate to monitoring” is not fully achieved.

    The second point that I would like to expand is the proposal to remember that there are many nuances. The most expensive of these nuances was worth repeating here again. For example, briefly list which Kubernetes parameters should be monitored.

    A good resume reinforces the main thoughts that viewers should take out of the speech, and in this case it could be done more effectively.

    Detective Order

    Pay attention to the story “Who ate the whole cluster”, which starts at 45:10 . The purpose of the story, in fact, is to show how drill down works, to bring us to the scheme from slide 326:

    But it would be boring to directly show this scheme, and Dmitry tells us the story of the investigation of the problem on the cluster, in which this scheme appears naturally. The meaning of the reception in this case is that we tell the audience some fascinating practical situation, and the main material that I really want to convey is the design, the surroundings for this situation. So you can achieve much more attention to the audience, which Dmitry successfully does.

    The ending is a little oiled: the butler-killer was never found, what happened to the Redis master was left overs. But to the right point, the audience kept their attention, and this is important.



    I already mentioned that there are 344 slides in Dmitry’s report, and now it’s a good time to talk about that there’s nothing wrong with that. To begin with, slide is a purely technical concept. You can speak with materials in PowerPoint, Keynote, Google presentations or any other tool that allows animations, or you can convert it all to .pdf, taking each change to a new page. The visual range does not change, and the formal number of slides is growing dramatically.

    Surprisingly many people have survived to this day who measure time by the number of slides, without thinking about what is depicted on these slides. They are afraid to break their complex slides into sequences, because it seems to them that the timing of the report will grow from this, they are used to the “minute on slide” and do not want to backtrack on it. Dmitry's performance is an example of how timing will remain under control. Only maybe the finger on the clicker gets tired.

    Instead of the technical concept of “slide,” I would suggest using the semantic concept of “frame”: one picture, one visual context. For example, such a frame

    is built on the screen on seven consecutive slides (30 - 36), and takes a minute and fourteen seconds in time starting at 05:05. This is a very important, fundamental frame for the report, the rest are mostly faster (Dmitry’s frames are still much more than minutes in the story).

    This approach is indispensable for demonstrating complex schemes. For example, slides 296 to 304 perfectly show the meaning of the architecture depicted on them, precisely because of the dynamics.

    I urge everyone to take an example from the speaker and no doubt add new elements of information to the slide through a click, and not immediately show the final picture. It’s much easier to follow the speech: the speaker always talks about the last object that appeared on the screen. It doesn’t matter if you use animation or create a new slide for each change, do as you prefer.

    With this approach, there is one technical difficulty that needs to be remembered: Acrobat Reader, a software for demonstrating .pdf slides at many conferences, does not have a preview mode for the next slide. That is, all of their three hundred and odd clicks must be remembered in order to synchronize speech and video. It looks rather sad when the presenter says something under a static picture, and then he catches himself and clicks on the slides with the comments: “so, I already told you, this also told ...”

    Accuracy and alignment

    If you look at the previous picture again (slide 36), you will notice that the circles symbolizing the elements and the properties of good monitoring are not evenly laid out on it. It seems that there is some idea in this: in the upper row they are larger with a gray outline, and on average they are smaller with a black one, so the coordinates can be not accidental. But I can’t understand this plan, and in my head at such moments unproductive thoughts of the form arise: did the author want to tell me something or did he just lose his sight?

    This is not the only case, there are others with a similar problem, for example, slide 76:

    In addition to aligning the circles, there is also a different font in the left and right parts of the screen. Is this a bug or a feature?

    Using Images

    The main rule for working with pictures in a presentation is this: if the image is the whole point of the slide, it should also occupy the entire slide. Everything is very clear here, nothing superfluous:

    But what can be improved is the unity of image style. It looks unnatural when photographic pictures are side by side with drawn-cartoon ones. The intersection of two worlds causes the viewer discomfort, although it is not always easy to formulate.

    Who Framed Roger Rabbit?


    It seems to me that the secret to the success and high ratings of Dmitry's reports is as follows:
    • The story develops very dynamically, something happens all the time.
    • A lot is illustrated by examples, stories and analysis of specific practical situations.
    • Visual materials are also dynamic, complement , and do not struggle with speech, and are well synchronized with it.

    Well, the topic is relevant, what is already there.

    Regular reviews

    If you want to receive feedback on your presentation, then I will gladly provide it to you. It happens that a job fills me for a long time or for some reason there is a break for several months, but progress, as we know, will not stop.

    What is needed for this?
    • Link to the video of the speech.
    • Link to slides.
    • Application from the author. Without the consent of the speaker himself, we will not analyze anything.

    All this needs to be sent to the habruiser p0b0rchy , that is, to me, Roman Poborchem. I promise that the review will be constructive and polite, and will also highlight positive aspects, and not just what needs to be improved.

    The most effective thing you can do to learn how to perform cool is to perform.

    Therefore, my friends, I invite you to submit an application for HighLoad ++ 2018. Deadline September 1, there’s no way to think about the topic for a long time, but there will be time to prepare the report. This I promise you, as well as tips, analysis, runs, and everything you need to make it cool.

    And subscribe to the newsletter for speakers (where all such materials regularly get).

    Also popular now: