Exam AZ-104 topic 6 question 15 discussion

so to just explain a bit better hopefully :) look at the top right you can see it is auto updated every 6 hours so within 24hours it is checked 4 times (4*6=24h) . It would need to scale up to have a bigger CPU to support the load that is getting as it is currently 100%

Answer is correct Box 1 : look at the top right of the picture it says 6 hours so 24hours/6hours = 4 times box2: this is app plan and VM so you scale up only

4 times Scale out: This means adding more instances to distribute the load Scale up :This means increasing the resources (CPU, memory) of the existing instances. => Scale out will be a better option

I don't see why people are saying scale up/out. It's averaging 3% cpu usage.

In real life, seeing the usage report, I would honestly scale the app down. It's barely using its CPU resources except the occasional spikes - probably because some job is running at that time. To save on costs and have better resource optimisation, I would scale it down. The exception to this rule would be that if during peak times, when the CPU is at 100%, the application is having performance issues that affect end users or causes whatever job runs on it to fail. If not, then I don't really care if the CPU peaks 100% and it would be better to have constant usage, let's say in the 50-60% on average with the occasional 100% spikes than just keeping the CPU almost idle for the majority of the time. If this comes in my exam I will answer 4 times and scale down.

in exam 26/12/2023

question 15 Box1: 4 Box2: scale up app service plan can scale up and scale out depending on the app service tier, i assume this is shared compute tier as there is no mention of it being premium or isolated tier in the question. unless i missed some detail. however based on the exhibit it shows cpu is 100% most of the time, which in my openion we should tier up and scale out.

I would scale down... maxes only matter if its causing issues and you can tell by the avg it isn't an issue. We do this stuff every single year. I would get no support to scale this app if it was performing as such. This thing could be hitting 100 just on its boot 1x day.

Passed my exam on 15 OCT. This question was in the exam. Thanks ET Team.

Mlantonis we desperately need you

I took the exam around 14th of Aug, this question came out... too bad I did not go thru this question before the exam

Microsoft will really like your answers on theh second box, but dudes, really, would you really scale up or out an App service that only has less than 4% average CPU utilization ? Sizing resources based on the Max CPU maybe would be OK for a really latency-critical application, but for most "human" scenarios I would actually scale down. Happy to be disrpoved, but maybe I'm just used to work in companies that are more attentive to costs than you guys here ;-)

Box 2: Scale Up. https://azure.microsoft.com/en-au/resources/cloud-computing-dictionary/scaling-out-vs-scaling-up/ Scale up when: 1. You see that your workloads are hitting some performance limit such as CPU or I/O limits. 2. You need to quickly react to fix performance issues that can't be solved with classic database optimization. 3. You need a solution that allows you to change service tiers to adapt to changing latency requirements. Scale out when: 1. You have geo-distributed applications where every app should access part of the data in the region. Each app will access only the shard associated to that region without affecting other shards. 2. You have a global sharding scenario—such as load balancing—where you have a large number of geo-distributed clients that insert data in their own dedicated shards. 3. You've maxed out your performance requirements, even in the highest performance tiers of your service, or if your data cannot fit into a single database.

Four times, scaled up - First observation - 30 days - 30 peaks (roughly) in average graph. So focus on a single day - how max cpu graph can be averaged over a time window to get the average graph. A 1 hr window would vary rapidly, a 24 hour window would be smooth as silk - a 6 hr window would give the current smoothness of the average graph - so average CPU calculated 4 times per day. From max graph, it looks like the web app is going through CPU deprivation so a scale up would definitely help alleviate the current issues with performance. Honestly - a 6 hour moving window for average graph would be calculated much more than 4 times a day, but it is the best answer from given data.

The second box, I would say: scale down: Avg CPU is only 4% (occasional spikes of 100% don' t matter that much).

Box2 should be Scaled Out. The average CPU is <10%, and we see occasional and continuous spikes over 70%. This says that most of the time the hardware is barely used (<10%), and we need to autoscale when there is a heavy load (>70%). Scaling up to a new SKU (scaling vertically) has a limit and is kind of old school and a waste of money during inactivity (<10%). Azure gives us the ability to scale out to more infrastructure when needed and scale in when the infrastructure is not in demand.

- the chart shows 6 hrs interval. so 24/6 = 4 times. - for CPU you need to scale UP.