long story short. Average is a poor measure of non linear phenomenon like the measure of a performance for a task .
Histogram are better to see the repartitions of values, and can hint you on a n-modal behaviour. (a «mode» when everything goes well, and a degraded one for instance which can hint on a potential transition of behavior).
I guess the next step is to say this is even more important to know the evolution over time and announce that the equivalent of spectrogram (3D mapped on 2D plus colour) is a good idea to known towards which mode you are tending the more. [EDIT: which can also more significantly graphed as spectrum analysis, colour encoding the amplitude of measure for a given frequency, and frequency being 1/Period thus making sure that slow processes may appear between 0 and average]
A shortcut is to hire of of the many jobless guys with a master in science that actually have experience in labs.
Beware that for a reason that is beyond my reach, people with background in math, management, literature have problems with scientific concepts that has been discovered after 1800 unless it is packaged like magic.
Replace the average (mean) with the median and you will have better results in almost all cases.
When the data is good the result ist virtually identical. When there are outliers, the median just ignores them unless there lots of them. When the data is that bad, that the median doesn't work anymore you need to up your statistics quite a few notches.
In psychophysics we either use the median for reaction times (Gamma distribution) or we normalize with something like the logarithm.
I think some of the techniques might be applicable to system performance analysis, so I submit the following introductory publication, in case any are interested [0].
I'd like to agree about scientific experience, but it's tragic how much trouble many professional scientists, or ex-scientists, seem to have with anything outside their specialism, like quantitative analysis of their computational techniques. When did you last see error bars on a benchmark, even?
SystemTap enables the same kind of analysis for a few years already. And it works well with older kernels (starting from 3.2, but really from 3.13). Of course, an in-kernel solution is better (and I believe that SystemTap will soon use it).
It's unfortunate that SystemTap never had much visbility. It also comes with a huge libraries of tapsets.
Systemtap was fantastic, but super finnicky. You had to have debuginfo for the kernel installed plus the source. This was fine for RHEL but kind of a pain elsewhere. Then you run into other problems like sometimes the distro compiler wasn't used to build the kernel you are running and suddenly you couldn't load any stap script. Also the only way to get output was through stdout, there was no easy way to programmatically pull results from an stap script.
BPF/BCC solves most of this. You still have to have the sources so you can get the right targets for kprobes, but you don't have the compiler mismatch problem. You can now access the hash maps directly from user space so you can do things like build system monitoring tools that run in production and record really specific information easily. I used systemtap for years, but bcc/bpf is a whole new world.
And that last sentence makes it obvious why he's using the past tense. Systemtap is in his past. When it was last committed to is irrelevant.
If I was feeling less gracious, I might call into question whether you actually have no horse in this race, as apparently calling other people's motives and statements into question is apparently fair game for this thread.
It might have been clear had it been used to set the tone. But used like this I think it deserves criticism whether intentional or not.
While I certainly can be be wrong, I also wasn't assured it wasn't intentional by the way it was explained. I suggest we all mind phrasing and get along this way.
P.S. I don't find it gracious of you to imply that I might be a lying cheat.
> I don't find it gracious of you to imply that I might be a lying cheat.
I wasn't. I was using the idea that someone could imply that to point out your own behavior. That you took it as negatively as you did and assumed there was some deeper implication was the point (that is, that you could, not that it was intended). The only thing you didn't do is connect it to your own behavior in your prior comment.
I never denied that I have made an implication. I believe I had better grounds for it (or I wouldn't have invested time in checking the project's state in the first place), provided a clearer point for OP to counter, and that that my tone was not as provokative.
If you can imagine a more tactful yet still effective alternative deterrent, I'm open to suggestions. I don't believe that just pointing out the fact of the use of the inappropriate tense would have been enough.
> I never denied that I have made an implication. I believe I had better grounds for it
I think you interpreted the statement a specific way, and went about your response without considering that your interpretation may not have been their intent. If you had allowed for the possibility of misinterpretation, I imagine you might have been a bit less aggressive with the accusation initially.
> that that my tone was not as provokative.
My view of your tone was informed by your assertion of the intention of another commenter. I generally find it to be bad form to tell someone (correct them on) their intention, rather than your interpretation of their intention.
To clarify, stated as fact "It's not a typo" means you reject the original commenter's possible explanation, but since we're talking about their intent, that is in effect calling them a liar (since they are the canonical source of knowledge as to their intention).
Perhaps that wasn't your intention, and the downvotes you received were unwarranted. If so, you have my apologies, and my suggestion that you be careful about using absolutist terminology when referring to people's intentions when it contradicts their own assertions without supplying good evidence (in this case, the evidence you supplied did not negate other explanations). There's a high bar to adequately explaining someone is wrong about their professed reasoning, rightly so.
> If you can imagine a more tactful yet still effective alternative deterrent, I'm open to suggestions.
Originally:
"I'm confused as to why you are using the past tense for SystemTap, the project is still very active."
In the follow up:
"Are you sure typo is what you meant? Your last sentence seems to indicate it's in your past as well." (If you don't see that last sentence for what it is, an indication that SystemTap is in this person't past, and thus they refer to it in the past tense to relate their experience to some former version of it. This is useful to notify others that your information on the topic may not be completely current.)
I did consider it, as evidenced by the words "seems" and "might". As for typo - it was not it, absolutely. Note that this fact doesn't necessarily imply dishonesty, but it is a possibility. This other comment was harsher indeed but likewise for a reason. The way I see it, it seems probable that OP was deceptive and my intention was to call them out in case it was. I'm still unconvinced by the excuse. Your suggestion of how I should have handled it is exactly the option I considered inadequate. I've addressed the last sentence and don't care to address the original comment yet again. This is one of the longest threads I've been a party to on here, and no offense to you, but I'm not sure it was worth it.
> I did consider it, as evidenced by the words "seems" and "might".
Your "seems" and "might" were modifiers on him "having a horse in this race." Unless your comments were meant to be taken entirely separately, the implication is "you are using the past tense because you have some vested interest in one of the projects." That's a fairly heavy accusation.
> The way I see it, it seems probable that OP was deceptive ... Your suggestion of how I should have handled it is exactly the option I considered inadequate.
See, I think this is one of those instances where your state of mind regarding this is coloring your interpretation of events. If you think that original comment using past tense was meant to imply the project was dead, I don't see that. To me that seems far less likely than the person used past tense because their experience was in the past, because I have done that myself, I see it on a regular basis, and I don't see people using it to imply a status of projects that doesn't match reality. Have you encountered this before?
> This is one of the longest threads I've been a party to on here, and no offense to you, but I'm not sure it was worth it.
Sure, I understand that. I'm responding because you're being civil in a discussion that's centered around a supposed mistake of yours, and I find understanding people's reasoning and state of mind when communicating to be somewhat interesting. Don't feel obligated to continue for my sake though.
You're forcing me to respond. You can't just remove my modifiers and then claim "heavy accusations".
Note also the irony of supposing someone's state of mind while advocating proof.
You've already shared your opinion on the past tense and I've disagreed. I'll add to that the use of third person instead of the first. Hopefully this will be my last referral to the OC.
I'm simply claiming that it is probable that it was an attempt at manipulation and that my reaction had reason and purpose outweighing it's "harshness".
I don't want to continue this if you aren't getting anything out of it. My intent isn't to agitate.
> You can't just remove my modifiers and then claim "heavy accusations".
I'm not sure where I removed the modifiers? Even when tempered by modifiers, and accusation of being wrong required a different level of evidence than of wrongdoing, in my opinion. E.g. "I think you might be incorrect" compared to "I think you might be trying to deceive us". Even mentioning that you think someone is not acting in good faith should be reserved unless you are fairly certain (as outlined by the guidelines). Of course, as you've stated, you felt this situation was at that level. I think most people looking in did not, which is where the friction came from.
> Note also the irony of supposing someone's state of mind while advocating proof.
I don't think it's ironic to suppose a state of mind when the purpose is to give a benefit of a doubt. It's rude, and not conducive to discussion to assume negatively (and then act on it), I don't think the same can be said for asking for keeping an open mind and asking for clarification (without an accusation, or with one but making it very clear you are looking for an alternative explanation).
> You've already shared your opinion on the past tense and I've disagreed. I'll add to that the use of third person instead of the first. Hopefully this will be my last referral to the OC.
> I'm simply claiming that it is probable that it was an attempt at manipulation and that my reaction had reason and purpose outweighing it's "harshness".
Fair enough. I was just trying to understand how you weighted the probabilities of what you thought the commenter was trying to convey to make you think that was the probable intent, because we obviously interpreted it very differently. If you don't want to go into your reasoning, I'm not sure there's much else to say.
> You removed the modifiers when you were paraphrasing.
Because I was talking about the implication itself, not the whole sentence. The implication was clear.
> Obviously to explain was my exact intention, which is why I provided several arguments to your one.
You're convinced of the intent of the original commenter, I have doubts. It's natural that I don't need to provide as many arguments, I just need to justify enough reasonable doubt in your mind as to your assessment of the original comment to make you question whether your approach was appropriate (for whatever level that is for you).
The crux of this is your assessment of the original comment, but you're unwilling to revisit that. I've never experienced what you seem to think was going on here (IIUC, the use of past tense to imply an incorrect status of some project or person for the purpose increasing the estimation of the thing it is actively being compared to). That obviously colors whether I think that's a likely scenario for what we had. When pressed if you have experienced something like this before, or for any justification of why you think this rare (in my estimation) scenario is what we witnessed, you've decided you don't want to continue down that line of inquiry. That's your right, but I don't feel like there's much useful information to mine here with that avenue cut off, as I view it as the most relevant.
> Honestly, I find your first sentence hard to believe.
Yet you keep responding. I'll make it easy for you, since by that statement you've left me to conclude two likely situations. Either you really don't want to be part of this conversation but can't help yourself, or you think I'm not conversing in good conscience and am instead trying to troll you. Neither situation is one in which I feel comfortable continuing, so this will be my last reply in this thread. The best thing I can do to actually communicate my intent is to follow through on my statement that my intent is not to agitate.
Feel free to respond if you like, I'll be notified if you do and will read your response, so you have a chance to express any final thoughts you like on this issue or the conversation in general.
Yes, and I've given SystemTap plenty of chances and hours of my time. But right now let's give BPF a chance -- it's more in-kernel in 2 years than SystemTap did in 10 years.
> It's unfortunate that SystemTap never had much visbility.
Right. DTrace was marketed. I don't think SystemTap was.
Is SystemTap stable already? As far as I remember from few years ago, it could easily crash your kernel with a valid program that didn't abuse anything, what disqualified it from use in production.
There is something important about this post (even though it looks like marketing): it's the first application vendor to create BPF scripts. It's a milestone for BPF. It won't be the last.
Fortunately in technology you can have both a high quality informing and informed post that is also sponsored. Posts of the nature "this is a tool we use and how we use it" are informative even if they are from the perspective of a company.
You only have to write the kernel stuff in C (which is a small part, since you are just collecting data), for the front end you can use BCC (Python/Lua)
Yes, we could, and I imagine we will at some point. It'll have become a user-level problem: the kernel capabilities are there (BPF+everything else), and there's the bcc user-level libraries (written in C) that make using the kernel stuff easier. The easiest way to add a front end would be to consume the bcc libraries. There's plenty of examples of how to use them (the Python and lua interfaces).
So if a scripting interface existed, would you use it? For what? Yes, it'd be nice to have, but I think a lot of people will get as far as using the provided bcc tools and be happy with that.
If we could write bpf programs in something other than C, it would go a long way towards opening up eBPF to people who wouldn't be able to write their own scripts otherwise.
Histogram are better to see the repartitions of values, and can hint you on a n-modal behaviour. (a «mode» when everything goes well, and a degraded one for instance which can hint on a potential transition of behavior).
I guess the next step is to say this is even more important to know the evolution over time and announce that the equivalent of spectrogram (3D mapped on 2D plus colour) is a good idea to known towards which mode you are tending the more. [EDIT: which can also more significantly graphed as spectrum analysis, colour encoding the amplitude of measure for a given frequency, and frequency being 1/Period thus making sure that slow processes may appear between 0 and average]
A shortcut is to hire of of the many jobless guys with a master in science that actually have experience in labs.
Beware that for a reason that is beyond my reach, people with background in math, management, literature have problems with scientific concepts that has been discovered after 1800 unless it is packaged like magic.