locked
Azure Scale Out , how does it actually work , help please ? RRS feed

  • Question

  • User1253338400 posted

    Hi,

    My app service runs a task every minute , so a new thread is spun up every minute. Now each thread needs to process hundreds of transactions.

    Ive scaled out using Custom Autoscale. So the metric i used was CPU above 50% to increase instance by 2 with a max of 10. Ithen have a scale down once it is below 30%.

    What i dont understand is:

    How can i see the insatcnes increase as the service is running and then see it scale back down ?

    Does each new instances that is created consume its own set of memory or does it share the memory of the server, so if the server has a max capacity of 8g, then there will be 3 instances and each will use approx 2g ?

    If the initial instance has gone over 50% , 2 new instances will be spun up , I assume the workload on the first instance is runninng and when the next task is spun up it will be allocated to the second instance , is that correct ?

    Does each new instance created get its own ip address in a specific range ?

    So is the creation of a new instance just like increasing specs of the machine ?

     Thanks

    Tuesday, January 26, 2021 11:46 PM

Answers

  • User-474980206 posted

    It’s depends on the technology of the service. Docker containers are a common approach, so to scale up a container is started. With some a new vm is created. some are just workloads. The azure team is always working on this to make scale out startup quicker.

    in practice it a new environment with its own memory, ipaddress and resources. To monitor see

      https://docs.microsoft.com/en-us/azure/azure-monitor/platform/autoscale-common-metrics

     

    • Marked as answer by Anonymous Thursday, October 7, 2021 12:00 AM
    Tuesday, February 9, 2021 3:58 PM