Hosting JSS in AWS

[amarks]

Hello,

Our company initially started with JAMF cloud, due to a variety of issue we want to host our own instance in AWS.

We want to do a dual instance set up with packages stored in the cloud.

Can someone give me guidance on the VM specifications for the cloud? I am planning to have two thousand users enrolled to start.

I had requested two vm's 16GB/500GB 2CPU's to start, but i am getting some resistance with this.

What's a good rule of thumb with this? Thanks for any guidance you can provide.

Andrew

[chriscollins]

Really, your specifications should generally match what they'd be self hosted in your own data center.

We somewhat recently moved our own data center based JSS's to the cloud. We have three AWS EC2 instances. The instance for the databaseis a 16GB/RAM and a 200GB hard drive (which with our database size actually is quite overkill). We have our primary public facing Tomcat instance with 8GB/RAM and a much smaller hard drive. Just enough to safely handle logs and log rotation.

We have an even smaller internal facing only instance that is "off to the side" althatbis used for access to the admin console. 4GB/RAM, 30GB hard drive.

We could get away with something smaller if we really wanted but we like having things split up.

For package distribution we use BOX but you could use S3.

[davidacland]

The HDD size isn't so much of an issue, although you may want to separate out the MySQL database location and backups so if they suddenly grow, they don't fill the server boot drive.

RAM is a big consideration with the number and duration of the https requests being one of the main factors and will depend on the check-in frequency, number of policies and their frequency, and how much you're using Self Service (as it keeps a connection open). I would say 16GB is as low as you would want to go.

I've had cases where 1 CPU (running linux) is ok, but in most cases, 2 is needed as a starting point.

Other consideration is whether you would want 1 MySQL server, 2 x tomcat servers and a load balancer between them for resilience. That is probably how I would do it.

[bmarks]

You should ask your account manager for the scaling doc. They may tell you it doesn't exist at first, but it definitely does.