Microsoft Fabric Capacity Smoothing and Data Warehouse Throttling

Поділитися
Вставка
  • Опубліковано 5 вер 2024
  • In the dynamic world of technology, demand spikes and troughs are inevitable. In this video, learn more about capacity smoothing that intelligently distributes consumption, mitigating the impact of sudden spikes and ensuring consistent performance. 
    Are you also curious to understand the admin and end user experience when you overload the capacity you have purchased? This video will uncover the below: 
    ✅ Explain the benefits of capacity smoothing 
    ✅ Explain capacity throttling principles  
    ✅ Demos on capacity admin and end-user experience 
    🎙 Meet the Speakers:
    👤 Stijn Wynants: Senior Product Manager for Fabric Warehouse
    Linkedin:   / stijn-wynants-ba528660  
    👤Sowmya Sivaraman: Senior Product Manager for Fabric Warehouse
    LinkedIn: Sowmya Sivaraman | LinkedIn
    👤Jonathan Garriss: Senior Product Manager in Fabric
    LinkedIn: Jonathan Garriss | LinkedIn

КОМЕНТАРІ • 12

  • @i.k.986
    @i.k.986 3 місяці тому +1

    Thank you for this clear explanations!

  • @carloscantu75
    @carloscantu75 9 місяців тому

    Very clear and useful explanation, thanks for sharing!

  • @TonioLora
    @TonioLora 9 місяців тому

    Great explainer.

  • @i.k.986
    @i.k.986 3 місяці тому

    maybe there are questions: what does the burndown do? The smoothing takes place always, I mean at least for specific activities, right? When the capacity is turned off, and activities are smoothed, during the off period of the capacity, the capacity is still charged, right?

  • @rodelharvey
    @rodelharvey 6 місяців тому

    This is great video. Looking at our timepoint. What does that mean if we have 7680 CU and for the particular timepoint we used 8000 CU. what is the impact? how do we calculate that 10minutes and how we find out if we are still good with 10 minutes?

  • @DanielWeikert
    @DanielWeikert 9 місяців тому +1

    Can you elaborate on the following. We have Power BI Premium capacity. I was informed we could use it to handle fabric workloads? Is that true. Can you elaborate on that. Thx

    • @sowmyasivaraman7731
      @sowmyasivaraman7731 9 місяців тому +1

      Yes, that is accurate. If you have P or F SKUs you will get the full Fabric functionality as long as your admin hasn't disabled Fabric workloads.

  • @tomfisher1099
    @tomfisher1099 9 місяців тому +1

    at the 4:20 mark you start to explain - if you account is 10 units, and when you've got 10 minutes of 10 units you are over capacity - what does that mean? If I have 10 units, can't I used 10 units a minute for 10 minutes? Or is that 1 unit per minute overage is then 10 total minutes over?

    • @jonathangarriss8878
      @jonathangarriss8878 9 місяців тому +1

      Glad you asked so I can help clarify this for you. Fabric allows you to 'burst' which means you use more capacity than you purchased. In your question if you bought 10CUs then yes you get to use 10CUs per second/minute/hour/day... But if you use 20CUs you start to eat capacity from the future. The table I'm introducing at that point in the video shows how we throttle based on how much future capacity you have consumed. So, if you buy 10 units and then use 10 units we let that go without doing anything to your account for ten minutes. But if you consume 20 units on that 10 unit capacity for >10 minutes we start delaying your new requests by 20seconds (AKA throttling you). If you use 110 Units in one minute you will also have used the next ten minutes and your account will go into throttling delays. I hope this helps.

  • @DanielWillen
    @DanielWillen 9 місяців тому

    Can we see the capacity usage? Which queries use most capacity and such?

    • @sowmyasivaraman7731
      @sowmyasivaraman7731 9 місяців тому

      Good question. Yes, you can see that through timepoint drill through page in metrics app. It will display the queries that ran (or were getting smoothed) at that timepoint, and also display the %capacity consumed, capacity units, identifier of the query etc. Hope this helps.

  • @v--p
    @v--p 7 місяців тому +1

    This is crazy, I can understand things running slow when you're over capacity or rejecting refreshes but actively rejecting interactive requests is called an outage. Try to explain that to the user base why they can't open reports. Also looks like usage is summed up over the 24 hour period and not what's currently running. So if I had a refresh that runs every 3 hours and takes 25 percent of capacity during the run, if that refresh runs 4 time a day I will be at 100% usage on capacity.