Denis Rystsov

@rystsov

Hooked on distributed systems. It's a personal account and I don't represent my employer.

Seattle, WA
rystsov.info
Joined December 2008
72 Photos and videos Photos and videos

Tweets

You blocked @rystsov

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @rystsov

  1. Pinned Tweet
    Feb 22

    My paper on CASPaxos, a simpler than Raft & Multi-Paxos protocol for building replicated state machines, is now on arXiv! The protocol has less moving parts, requires less code to implement, supports membership change and has formal proof -

    19 replies 280 retweets 666 likes
    Show this thread
    Show this thread
    Undo
  2. Retweeted
    Aug 23

    We worked with to analyze their distributed graph database, and found 23 issues, including crashes, deadlocks, consistency anomalies, and data loss. Dgraph has addressed almost all of these in recent releases, though some problems remain.

    41 retweets 109 likes
    Undo
  3. Retweeted
    Aug 9

    Surgeons didn't want to use checklists because they were too full of themselves, but then accidental deaths fell by 30-50% in hospitals that adopted them. Know who else often suffers from the same hubris? Programmers.

    17 replies 535 retweets 1,495 likes
    Show this thread
    Show this thread
    Undo
  4. Retweeted
    Aug 7

    My sabbatical at Azure Cosmos DB

    21 retweets 42 likes
    Undo
  5. Retweeted
    Jul 16

    Oh god, this bullshit again. There is ZERO evidence that any type checker less powerful than Coq will catch more bugs than being healthy and well-rested. If you use Haskell but eat junk food and don't exercise, then you're unethical.

    Programming without a type checker is unethical.
    14 replies 101 retweets 360 likes
    Show this thread
    Show this thread
    Undo
  6. Jul 12

    Wow, "Testing shared memories" paper is a gem. Last night I combined an idea from the paper with Jepsen and the complexity of checking linearizability dropped from O(n!) to O(n). It made possible to run consistency checking experiments for hours

    2 replies 21 retweets 97 likes
    Undo
  7. Retweeted
    Jul 8

    Moscow is great. Sochi was sweet. Rostov was interesting. But St.Petersburg. St.Petersburg is the woman that steal your heart, is the love of your life, is the burning sky where you want to disappear in

    72 replies 1,248 retweets 4,738 likes
    Undo
  8. Jun 27

    I did the same, chose treap and nailed the interview 😆

    Interviewer: hmmm Can you think of any way we can make this faster? Me: Sure, lets replace this array with a log-linear histogram Interviewer: Uh.. Me: Sit the fuck down, your smug ass wanted to play fancypants with the datastructures. Now we're here for the night
    Show this thread
    1 retweet 2 likes
    Undo
  9. Retweeted
    Jun 22

    If I’ve learned anything from watching really amazing metal / wood craft videos on YouTube, any legit high complexity projects spend easily 50-90% of their time building tools to support high precision sub tasks. Or something. The figure is fictional, but the pattern real.

    2 replies 5 retweets 26 likes
    Undo
  10. Retweeted
    May 31
    Replying to

    At some point I'm gonna have to figure out what language to use for this because vendors have gotten enthusiastic about saying "passed" without discussing how the system also failed, what was tested, and limits of Jepsen's scope

    2 replies 2 retweets 4 likes
    Undo
  11. Retweeted
    May 9

    One of the most compelling benefits of CASPaxos is in storage. Since logs aren't reified, you never need to write them to disk - less write amplification. Since registers are independent, you can batch & write out of order - high concurrency disk ops

    2 replies 2 retweets 15 likes
    Show this thread
    Show this thread
    Undo
  12. May 9

    Devirtualized with he’s cool. Transactions, Australia, Cosmos DB, Orleans, CASPaxos, Russia, consistent backups, USSR, China, education, freedom, fraud, Mardi Gras, urban geography, mechanical sympathy. It was hella good discussion!

    3 replies 1 retweet 21 likes
    Undo
  13. Retweeted
    May 1

    For quite some time I've been putting together a list of resources on testing distributed systems

    2 replies 21 retweets 44 likes
    Undo
  14. Apr 26

    By the way, we have positions including senior / principal roles. Ping me for details.

    3 replies 2 retweets 3 likes
    Show this thread
    Show this thread
    Undo
  15. Apr 26

    Hey, pals, I recently joined Cosmos DB, and it's an incredible place: teammates are constantly buzzing about TLA+, linearizability, performance, time, atomicity and other interesting stuff. The good news is we're hiring:

    3 replies 21 retweets 52 likes
    Show this thread
    Show this thread
    Undo
  16. Retweeted
    Feb 18

    Which logical fallacy you'll be doing today? 😎

    3 replies 43 retweets 80 likes
    Undo
  17. Retweeted
    Apr 5

    paxos made simple

    32 retweets 56 likes
    Undo
  18. Retweeted
    Apr 2

    Consistent Hashing: Algorithmic Tradeoffs

    2 replies 161 retweets 402 likes
    Undo
  19. Retweeted
    Mar 31

    my goofy talk about CASPaxos was put on the videos website, you can watch it here:

    4 replies 5 retweets 25 likes
    Undo
  20. Retweeted
    Mar 28

    Azure customers, we'd love your feedback on the organizational changes we're testing out on the Cosmos DB docs on .

    2 replies 32 retweets 27 likes
    Show this thread
    Show this thread
    Undo
  21. Retweeted
    Mar 27

    This is pretty interesting. Low coordination databases are something I've been thinking about recently, but coming at it from the linearizability camp. Still, what's to say a single system can't offer both? Both have their virtues and their shortcomings

    "Anna: a KVS for any scale" We et al., ICDE'18 Using coordination-free actors to build a key-value store that offers great single node and distributed performance.
    1 reply 1 retweet 8 likes
    Show this thread
    Show this thread
    Undo

@rystsov hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·