Heinrich Hartmann   

@heinrichhartman

IT Operations, Monitoring, Software Engineering, Data Science, Mathematics. HartmannIT /

Stemwede, Deutschland
Vrijeme pridruživanja: rujan 2012.

Tweetovi

Blokirali ste korisnika/cu @heinrichhartman

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @heinrichhartman

  1. Prikvačeni tweet

    We finally got around writing-up a formal treatment of the Circonus Log-Linear Histogram data-structure! Here is the TL;DR ... 1/x

    Prikaži ovu nit
    Poništi
  2. Finally solved this problem: AJAX request from Bookmarklet works just fine. Sometimes simple things are easy. CORS/CSP do not apply to AJAX requests. Just resource loading via <script> / eval(); which we don't use here.

    Poništi
  3. proslijedio/la je Tweet

    “If you downloaded your data, it may be incomplete, and it may contain videos that are not yours”. Google casually dropping that in there.

    Prikaži ovu nit
    Poništi
  4. Poništi
  5. proslijedio/la je Tweet
    28. sij

    Trying to get a better picture of our landscape via

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    11. pro 2019.

    The older I get^W^W^Wmore I read math textbooks, the more I recognize the parallels between the tension in writing clear code and the tension in proving theorems cleanly

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    2. velj

    Performance artist generates virtual traffic jams in Google Maps by pulling a wagon full of smartphones

    Poništi
  8. Poništi
  9. proslijedio/la je Tweet
    28. sij

    The paper on Circllhist log-linear histogram by & is published and can be found at . Including links to opensource implementations and code to reproduce their results.

    Poništi
  10. proslijedio/la je Tweet
    30. sij

    For those who missed it, I played with last night, expecting to snark all over it. Instead the experience was damned close to perfect, and now my opinion of has shifted as a result.

    Prikaži ovu nit
    Poništi
  11. "It's really hard to dehumanize, even the enemy" -- CIA & NSA director Hayden My take-away from is that we MUST NOT dehumanize anyone. Not "enemies". Not criminals. There are no "bad guys" -- only humans, with all their imperfections.

    Poništi
  12. proslijedio/la je Tweet
    28. sij

    I'm looking for someone to take over writing Monitoring Weekly. I'd prefer someone unaffiliated with a monitoring vendor (that objectivity is crucial!). You don't need to be an expert in the space, just a desire to become one. Let's chat. mike at duckbillgroup dot com

    Poništi
  13. proslijedio/la je Tweet
    28. sij

    This is an absolutely smashing paper on how the NHS hides data in plain sight causing anyone doing analysis (me included) no end of woe.

    Poništi
  14. - More deatils: Read the paper! - Try the Circllhist: - Reproduce the evaluation: We have source code & docker image on GitHub: fin.

    Prikaži ovu nit
    Poništi
  15. Finding 3: Circllhist + DDSketch + HDR Histogram deliver quantiles with high precision (<2% error) and have a-priori error bounds. - t-digest accuracy can tank after multiple merge setps - Typical Prometheus Histograms do not give competitive quantile accuracy (10 bins) 6/x

    Prikaži ovu nit
    Poništi
  16. Finding 2: Performance evaluation in our particular setup show Circllhist with best-in-class insertion + merge performance. t-digest is faster for quantile calculations. Source code: 5/x

    Prikaži ovu nit
    Poništi
  17. Finding 1: Everyone uses thier own definition of quantiles. We find type-1 quantiles make the most sense. In any case watch out when comparing them. 4/x

    Prikaži ovu nit
    Poništi
  18. For the evaluation we consider the following alternative latency summary technologies found in the wild: - HDR Histograms / - t-digest / - DDSketch = Log Histograms We also include histograms (not really fair comparison) 3/x

    Prikaži ovu nit
    Poništi
  19. The circllhist is a data-structure that allows to capture and analyze latency distribution. Here is how it looks. It's fully mergable, unbounded, fast (ns insertions), and reasonably small (<10kb) and accurate as well as battle tested for nearly a decade. 2/x

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·