Exam AZ-104 topic 4 question 87 discussion

2, 15 Initial instance is 1 as specified in first figure. 80% for 15 minutes reaches 10 minutes duration, but haven't reached second turn of scale out, so only one new instance is created. Since cool down time is 5 minutes, which means after one scale happens, it will count 5 minutes before counting a new 10 minutes, so 15 minutes total.

I don't know why but it seems that majority of people commenting here don't know exactly how autoscaling works. So let me explain few things. Duration value is only used for data range - how much data autoscaling system has to aggregate to determine if rule applies or not. Cooldown - how much time has to pass before next autoscale operation to trigger. So once you start you app the first autoscale may happen not earlier than after duration value (because you need specific data range). Each next will happen every cooldown value. So: - first scale out will happen after 10 minutes (duration value) - next scale out will happen after 15 minutes (+5 minutes of cooldown) Therefore answer for first question is 3 instances Second one is simpler. Scaling operation just happened. So next scaling may happen after cooldown time which is 5 minutes.

Run this question in ChatGPT to understand why the answer is correct: 2 and 15 "App1 is configured to auto scaling as such: Scale Out when CPU over 70% increase by 1 instance. Instances limit min 1, max 5, default 1. Autoscale rule criteria is: if CPU is more than 70%, for duration of 10 minutes, increase instances by 1. Cool down period is 5 minutes. The questions are: 1. How many instances of the app1 will be running if the CPU is 80% for 15 minutes? 2. After how many minutes another instance will be created after the first scale-out?"

Due to the lack of a clear answer, I created a lab. Here are the conditions: default VM count - 1, minimum VM - 1, maximum VM count - 5. VM size - Standard_B2s. If the CPU load exceeds 80% for 10 minutes, increase the count by 1; if it falls below 40% for 10 minutes, decrease the count by 1. Cooldown is set to 5 minutes in both cases. On the first instance, a script was run using PowerShell at 11:57, and the CPU load reached 95-100%. The first scaling occurred at 12:06, the second at 12:11, and the third at 12:17.

in first cycle - 10 minutes in second cycle - duration + cool downtime =15 mins

BOX1: 3 instances. CPU went up for 15min. it will meet the condition of greater than 75 for duration of 10min so 1initial +1. then Cooldown is 5min so after 5min it adds another instance since condition still 75. BOX2: 5 min many are confused here. after the scale out instance of 10min is met, it will rely on cool down time to increase the instance again and not the duration. 10min is just a condition. Since it's asking for the minimum time after the first scale out, 5min.

It seems to be 3 and 5 Answer (and test case as well) is here https://cloudbuild.co.uk/azure-virtual-machine-scale-set-duration-and-cool-down-explained/

6/13/24 on exam

This question was in exam 01/03/2024

Correct

First answer is 3 . After 10 mins, first scale out happens (+1), then after 5 mins of cooling the system checks if last 10 mins usage was above thresold, then 2nd scale out (+1). So total, 3 instances after 15 mins. 2. 5 , cooling period is 5 mins, so after first scale out, wait for 5 mins and then check again the usage.

2, 15 "The cooldown period for each rule dictates how long after the previous scale action (whatever rule [..] was [triggered]), the rule can be applied [again]." Source: https://github.com/MicrosoftDocs/azure-docs/issues/17169

In order to answer the question, is important to understand how Azure scale actions are triggered. Lets have same params as in question: 10 minutes duration of scale out verification and 5 minutes of cool down Once action meets criteria for 10 minutes, action is triggered. After that, 5 minutes is reserved for scale out action to take in place and App to rebalance the load. Next, if load is still meet the auto-scale rule, it will start counting another 10 minutes to see if even after previous scale-out load is high. If yes, only than a new instance will start running and another 5 minutes of cool down will be started. Therefore, answer for this questions are: 1) 2 instances; 2) 15 minutes.

The answers are 3 and 5. "The cooldown period for each rule dictates how long after the previous scale action (whatever rule [..] was [triggered]), the rule can be applied [again]." Source: https://github.com/MicrosoftDocs/azure-docs/issues/17169 The first question is a bit theoretical: it makes it appear that the load stays constant at 80% for 15 minutes - which would mean that after the seconde instance is created after 10 minutes, demand increases accordingly so that even then 80% load is maintained, with no load decrease. Anyway, were that to happen, then after another 5 minutes of cooling down, the rule would be allowed to be triggered again, after 15 minutes in total - to create a third instance.

OpenAI: Let's say you have an Azure Virtual Machine Scale Set that's serving a web application. You've set up autoscaling rules to add instances when the CPU usage exceeds 70% and to remove instances when it drops below 30%. After a scaling event, you've configured a 5-minute cooldown period. If the CPU usage goes above 70%, the autoscaler adds more instances to handle the increased load. During the next 5 minutes, even if the CPU usage remains high, the autoscaler won't take any further scaling actions due to the cooldown period. After the cooldown period elapses, the autoscaler will reevaluate the metrics and potentially trigger another scaling action if the conditions are still met.

Answer: 2/5 Question 1 - Some folks have done a good job of explaining but people still misunderstand cooldowns so I will try one more time, maybe a little differently. In our case the rule requires a straight 10 minutes of data aggregation above 70 to add another node. This does NOT mean that its additive in increments 10, 20, 30, 40, 50 and it scales out at each 10 minute interval. And it definitely does NOT mean 10+5 = 15. The 10 minutes is a SLIDING window based on the interval and the Cooldown Example |____over_70______| 10 minutes went by over 70 -- Scale up 1 |____over_70______|__STill_Over_70____| Cooldown of 5 went by and it was STILL above 70 so it scaled again. AKA in 15 minutes it scaled 2 times because it never dropped below 70 to reset the 10 minute aggregation need.

Answer is 2 and 5. Duration: The amount of time to look back for metrics. For example, timeWindow = "PT10M" means that every time autoscale runs, it queries metrics for the past 10 minutes. The time window allows your metrics to be normalized and avoids reacting to transient cooldown: The amount of time to wait after a scale operation before scaling again. For example, if cooldown = "PT10M", autoscale doesn't attempt to scale again for another 10 minutes. The cooldown is to allow the metrics to stabilize after the addition or removal of instances. https://learn.microsoft.com/en-us/azure/azure-monitor/autoscale/autoscale-understanding-settings So, if cooldown = 5 min and Duration = 10 min, after one scale (say at 10:15 AM), it will wait 5 min (10:20 AM) but it will look back 10 min which means it will look from 10:10 AM to 10:20 AM.