Azure Virtual Machines Contd

• Generally Applications will have peak hours where the need for resources would increase.
• In the on-premises systems, since we rely on hardware, we need to plan for the peak hour demands and arrange hardware accordingly.
• In the case of cloud, we can create virtual machines in matters of minutes any time and destroy them as well.
• So, lets take the case of
  • predicatable demand:
    • Example: Your application has a normal traffic of 100 users, but every day from 5-7 PM on an average you have 1000 parallel users
    • Now on cloud before 5 PM We can increase number of vms running our application to meet the peak demands and then destroy the machine
    • Normal times:
      
    • Peak times: We increase number of servers
      
  • unpredicatable demands: Number of users can be any number, so scaling to particular number of servers might not work out
• In all the clouds, there is an ability to automatically increase the number of servers (scaling up) based on some metrics and decrease the number of servers (scaling down), and this feature is called as elasticity.
• In the case of Azure we can acheive this by Virtual Machine Scale Sets.