How LinkedIn improved their latency by 60%

Sdílet
Vložit
  • čas přidán 27. 06. 2024
  • System Design for SDE-2, SDE-3, and above: arpitbhayani.me/masterclass
    System Design for Beginners: arpitbhayani.me/sys-design
    Redis Internals: arpitbhayani.me/redis
    Build Your Own Redis / DNS / BitTorrent / SQLite - with CodeCrafters.
    Sign up and get 40% off - app.codecrafters.io/join?via=...
    PostgreSQL per-client process model - • PostgreSQL connection ...
    PostgreSQL Internals Playlist - • PostgreSQL connection ...
    Database Engineering Playlist - • PostgreSQL connection ...
    System Design Playlist - • How DNS really works a...
    Other links
    CS Engineering and Software Development books that I have read
    arpitbhayani.me/bookshelf
    Research papers that I have read
    arpitbhayani.me/papershelf
    Newsletter: arpit.substack.com
    LinkedIn: / arpitbhayani
    Twitter: / arpit_bhayani
    Until next time, stay awesome :)
    Yours truly,
    Arpit
    arpitbhayani.me
    #AsliEngineering
  • Věda a technologie

Komentáře • 46

  • @NK-ju6ns
    @NK-ju6ns Před dnem +5

    It’s simple change but you make it sound like a rocket science.

  • @pratyushsinha3348
    @pratyushsinha3348 Před 2 dny +6

    Quality of content is insane, I have read limitation of JSON but never understood it as I did now. Big Thanks

  • @nagesh007
    @nagesh007 Před 2 dny +1

    Awesome 😍 , thanks for teaching

  • @tanaykmr
    @tanaykmr Před 2 dny

    A huge thank you for these videos sir!

  • @akash-kumar737
    @akash-kumar737 Před 2 dny

    Thanks 😊. Keep making such informative videos.

  • @akashshirale1927
    @akashshirale1927 Před 2 dny

    Thanksss a lot...Such quality content!!!

  • @saitrinathdubba
    @saitrinathdubba Před 2 dny +1

    Brilliantly explained !! just like protobuf, info is efficiently packed ;) Thanks a lot :)

  • @manikantapunnam1189

    Quality content, thanks for this ❤

  • @prathamkrishna3751
    @prathamkrishna3751 Před dnem

    Hi Arpit! Lovely video. Got a question for you, wrt the above example - "123456789" is consuming 9 bytes, and apart from that transferring data through json is a few more additional bits including the colon etc. Protobuf being one way of improving efficiency, but i believe csv would perhaps also be more efficient than json? We can eliminate the use of inverted commas, keys. Just return values seperated by commas and at the client end regex remove the comma. What do you think of this?

  • @iajaydandge
    @iajaydandge Před 2 dny +3

    Hi Arpit, thanks for informational video as always.
    I have read that protobuf are generally used in interservice communication and not browser facing APIs, because they don't provide significant improvements over JSON in javascript based environment. Am I right or wrong?

    • @AsliEngineering
      @AsliEngineering  Před 2 dny +3

      Yes, but this is not due to JSON. We do not see a substantial improvement because the latency of transmitting network over the internet is so high that cost of serialisation is miniscule in front of it.
      Hence, spending a few extra cycles does not matter much and hence Protobufs are typically preferred for inter-service communication within the private infra/vpc.

    • @iajaydandge
      @iajaydandge Před 2 dny +1

      @@AsliEngineering Thank you

    • @Shris-jc6eu
      @Shris-jc6eu Před dnem

      Good learning

  • @gagandhand8907
    @gagandhand8907 Před 15 hodinami

    Amaing. I have one question to ask Will serilization and deserialization not add any latency as client library will get the request in serialize format it needs to be deserialized.

  • @jatinnandwani6678
    @jatinnandwani6678 Před 2 dny

    Superb

  • @rajyobro2078
    @rajyobro2078 Před 2 dny +1

    Service mesh (istio/linkerd) solves this out of the box. Not to mention it’s language agnostic and decoupling service to service communication from app code. Library based framework is a thing of the past.

    • @SurendraKumar-tf2fc
      @SurendraKumar-tf2fc Před 2 dny +1

      As per my understanding, Service Mesh like Istio make sure the communication between services in such a way that it can be trackable. But, it doesn't reduce the response time and doesn't play any role in reducing the latency.
      LinkedIn might have not wanted to spin up more than the required pods for don't know how many appliations..😅
      To reduce the cloud cost, Organisation should also focus on the tech and library which they have been using at code level. We as a developer should not always rely on cloud infrastructure..🙂

    • @dew_reddit
      @dew_reddit Před 2 dny

      Marely managing & tracking traffic better doesn’t solve the performance problem.
      Also, linkedin is probably already using a properly configured service mesh.

  • @jst_kishan
    @jst_kishan Před 2 dny +1

    You mentioned that one of the possible solutions could have been to compress the data and transmit. But isn't the data compressed by default with gzip in HTTP communication?

  • @armaansaini271
    @armaansaini271 Před 2 dny

    How/Where do you find these information? These internal updates, and use cases by the companies? Is there any newsletter, website or channel or something?

    • @AsliEngineering
      @AsliEngineering  Před 2 dny +1

      nope. I regularly go through company's engineering blogs and publications; and also conferences like VLDB, SIGMOD, and USENIX.

    • @armaansaini271
      @armaansaini271 Před dnem

      @@AsliEngineering Thank you ✨

  • @yashagarwal8468
    @yashagarwal8468 Před 17 hodinami

    I was trying to check the api calls that my client makes when i open linkedin and i was not able to find any calls with protobuf. I'm curious if they still havent rolled out this feature in India, any ideas?

  • @nagesh007
    @nagesh007 Před 2 dny

    Code Crafters Recommendation from you is awesome and Mind blowing😍 .... Thanks a lot bro

  • @rajesh.singha
    @rajesh.singha Před 2 dny

    Good content

  • @rajchaurasiya1265
    @rajchaurasiya1265 Před 2 dny

    Kabhi kabhi lgta hai hmm software developer hai bhi ya nhi issko sun ne ke baad

  • @nirjalpaudel
    @nirjalpaudel Před 2 dny

    How do you find blogs like these and also papers

    • @AsliEngineering
      @AsliEngineering  Před 2 dny

      I regularly go through company's engineering blogs and publications; and also conferences like VLDB, SIGMOD, and USENIX.

  • @prashanthb6521
    @prashanthb6521 Před 2 dny

    Its surprising they didnt do this earlier and were paying high cloud costs till now.

    • @xiaoshen194
      @xiaoshen194 Před 2 dny +2

      They do excessive diversity hiring. No wonder…

  • @snehil3209
    @snehil3209 Před 2 dny

    Arpit you're true gem. You BILLIONS of subscribers ❤

  • @imchiragkumar
    @imchiragkumar Před 2 dny

    kanha reduce hua h, pura lag karte rahta h linkedin

    • @AsliEngineering
      @AsliEngineering  Před 2 dny +1

      If you look at their API calls, they are graphql based and have massive payloads. Takes a ton of time to transmit and process the data. So, still some scope for improvement.

  • @vishusingh008
    @vishusingh008 Před 2 dny

    BC ye video ke hisab latency kam ho gae hai, aur actual app me latency is all time high. The difference between theory and actual user experience 😅

  • @chobblegobbler6671
    @chobblegobbler6671 Před 2 dny +5

    Enough bro.. stop