Always-on Fleet Costing

Question

I have been unable to get any answers in how to model the rough user-cost for my deployment.  
  
I want an always-on fleet but am not clear what control I have over this.  
  
I want to delivery a peak usage of 500 users - but this will be very infrequent. The actual usage will probably be within a 8-10 hour window each day with the peak between 09:00-17:00 and very few users (less than 50) outside of these hours.  
  
Is it possible to use a combination of AWS-CLI and the Fleet scaling to control these and reduce costs?  
  
e.g. at 06:00 each day 100 instances spin up to support initial users and the fleet scaling options automatically add additional instances up to a maximum of 500 through the day. The fleet scaling also reduces down the number of 'on instances' based on active users, until 22:00 when all instances are shut-down (but still available to spin-up 'on demand' if requested?).  
  
This would mean that the cost-per user would be based on a combination of 'on-time' and 'off-time' throughout the day rather than cost for 500 users 'always on'.  
  
Can any one explain if this is possible/reasonable or how the costing works in practice?

Answer

Hi GrahamMacrae,   
  
There are two types of fleets in AppStream 2.0, Always-On and On-Demand.   
With Always-On fleets, all instances that are running are charged the applicable running instance fee, based on the instance type and size, even when users aren't connected.  
With On-Demand fleets, instances are charged the applicable running instance fee, based on the instance type and size, only when the instances are used for streaming sessions. Instances in On-Demand fleets that are not being used for streaming sessions are charged a small hourly stopped instance fee that is the same for all instance types and sizes.  
Another cost you will want to consider is User RDS SAL fees. For more information on pricing, please see: https://aws.amazon.com/appstream2/pricing/  
  
So controlling the costs by scaling your fleet can be done a few ways. Time based scaling is able to be accomplished via the AWS CLI, Link: https://docs.aws.amazon.com/appstream2/latest/developerguide/autoscaling.html#autoscaling-cli  
This can be used with other scaling methods such as Available Capacity or Capacity Utilization to ensure that the number of available instances stays inline with the number of active users you have throughout the day.   
  
Regards,   
  
Nate

Answer

You can find pretty much all you should need to know about scaling to your requirements on this web page:  
  
https://aws.amazon.com/blogs/compute/scaling-your-desktop-application-streams-with-amazon-appstream-2-0/  
  
It includes a tutorial for using the auto scaling as well as writing your Lambda function to scale based on time of day.

Always-on Fleet Costing

Relevant questions

Error! Cannot create the fleet. Reason: Current limit of fleets of 0 have been reached