Quick question. I imagine that nothing need to be done in terms of configuration this is applied automatically according to the Fabric Throttling "algorithm".
Qs: 1. Did I understand correctly that you get a fresh start every 24 hours even if smoothing can’t smooth all the used CU seconds over the 24 hour period? So in reality you face the same issue with the data scientist not being able to run loads every day? In short is smoothing carried over to next day? 2. Is there any additional costs related to bursting in the pay as you go model? 3. I understand that if you pause right after a heavy job then you will have to pay the extra CUs used. But does it mean that the expense of let’s say a F2 capacity can I crease above the monthly price as per the Microsoft pricing table? 4. In the reserved capacity prices is bursting and smoothing also applied? If not how do you handle the same situation you described in your video with the reserved model? 5. Is auto scale necessary when we have bursting?
Great questions. 1. Correct, there is an upper bounds for a 24 hour period of the CU that you can consume. When you hit it, then you will get messages on needing to scale up or out. This is why you may need auto-scale. It's most cost effective when you can buy the CU's you need in a 24 hour period. 2. Bursting is included in all capacities. No additional charge. 3. Yes, if you use all of an F2's CU in 1 hour, pause, then un-pause, use a full day of CU, and then pause again, you will have had 2x the spend of an F2 for the day. 4. Yes, bursting and smoothing is used in the reserve capacities. The 40% discount of reserve capacities along with bursting and smoothing make it very hard to pick the OnDemand capacities. 5. Yes, see answer 1 above. Please let me know if you have any follow up questions.
@@ChrisWagnerDatagod Thanks for the quick and well written reply Chris! I have some follow up questions: 1. So if I never pause, and choose the pay as you go model, I would never exceed the monthly cost as per Microsoft’s pricing table (except of course one lake storage costs)? 2. Doesn’t this make pausing a capacity very difficult? Like how do you know when to pause a capacity, so to not pay more than the price on the pricing table… 3. I am still a bit unsure about how over use affects let’s say a F2 capacity in terms of cost. If I use all CUs for the 24 hour in 1 hour, then pause, and then un pause and use fabric - wouldn’t it throttle and give me error messages because the rest 23 hours have been used? 4. How does bursting and smoothing work in the reserve capacities in terms of extra costs? If CU usage can’t be smoothed over a 24 hour period, do I pay for the extra CU use in that period or will I be throttled? I guess it’s not possible and doesn’t make sense to pause a reserved capacity.
Very useful, thanks for sharing!
Glad it was helpful!
Great Video ! Thanks for sharing !
Thanks for watching!
Very interesting. Thanks
Glad you enjoyed it
Quick question. I imagine that nothing need to be done in terms of configuration this is applied automatically according to the Fabric Throttling "algorithm".
Correct. This just works.
Is bursting and smoothing enabled by default? If not, is it turned on in Azure or Fabric Admin portal?
Good question. This is how the MWC works. No way to turn it on or off.
Qs:
1. Did I understand correctly that you get a fresh start every 24 hours even if smoothing can’t smooth all the used CU seconds over the 24 hour period? So in reality you face the same issue with the data scientist not being able to run loads every day? In short is smoothing carried over to next day?
2. Is there any additional costs related to bursting in the pay as you go model?
3. I understand that if you pause right after a heavy job then you will have to pay the extra CUs used. But does it mean that the expense of let’s say a F2 capacity can I crease above the monthly price as per the Microsoft pricing table?
4. In the reserved capacity prices is bursting and smoothing also applied? If not how do you handle the same situation you described in your video with the reserved model?
5. Is auto scale necessary when we have bursting?
Great questions.
1. Correct, there is an upper bounds for a 24 hour period of the CU that you can consume. When you hit it, then you will get messages on needing to scale up or out. This is why you may need auto-scale. It's most cost effective when you can buy the CU's you need in a 24 hour period.
2. Bursting is included in all capacities. No additional charge.
3. Yes, if you use all of an F2's CU in 1 hour, pause, then un-pause, use a full day of CU, and then pause again, you will have had 2x the spend of an F2 for the day.
4. Yes, bursting and smoothing is used in the reserve capacities. The 40% discount of reserve capacities along with bursting and smoothing make it very hard to pick the OnDemand capacities.
5. Yes, see answer 1 above.
Please let me know if you have any follow up questions.
@@ChrisWagnerDatagod Thanks for the quick and well written reply Chris!
I have some follow up questions:
1. So if I never pause, and choose the pay as you go model, I would never exceed the monthly cost as per Microsoft’s pricing table (except of course one lake storage costs)?
2. Doesn’t this make pausing a capacity very difficult? Like how do you know when to pause a capacity, so to not pay more than the price on the pricing table…
3. I am still a bit unsure about how
over use affects let’s say a F2 capacity in terms of cost. If I use all CUs for the 24 hour in 1 hour, then pause, and then un pause and use fabric - wouldn’t it throttle and give me error messages because the rest 23 hours have been used?
4. How does bursting and smoothing work in the reserve capacities in terms of extra costs? If CU usage can’t be smoothed over a 24 hour period, do I pay for the extra CU use in that period or will I be throttled? I guess it’s not possible and doesn’t make sense to pause a reserved capacity.
Why your title like 2 style of fart, lol
hahahha... that's what its called. Too funny.