Scaleset sizing

[Scott_Paisley]

Hi

If we want to implement a scaleset in Azure, is there a design guide or best practise around sizing? What I mean is, assume we need 32 CPU max throughput, should we build that as 4 machines with 8 CPU, or 8 machines with 4 CPU? Is it better to have more smaller boxes in the set or fewer bigger boxes?

Thanks

[MarcusJ · Accepted solution]

Both approaches are valid.

I would say it really comes down to if you have alot of big connection flows.
Otherwise you will still handle the same amount of throughput but on several VMs.

Bigger VMs = Can handle bigger connection flows
Smaller VMs = Can handle less big connection flows

But if i could choose myself i would go for bigger boxes, i have been running VMSS (Scale set) for past 6 years in West Europe region. My experience is that the more virtual machines you have the bigger is the chance that there is a problem with one of them as well that you can support bigger flows. I would say KEEP IT SIMPLE by choosing bigger boxes ❤️

For patching it doesnt really matter if you have big or small boxes, just deploy new VMs inside the scale set and remove the old member which is unpatched and the traffic will be changed from the load balancer.

But for licensing you can either buy per CPU from Check Point (Bring your own license) or use Pay as you go.
Generally Bring your own license is cheaper in the long run.

You need to ask yourself which benefits you want the most and which disadvantages you are ready to have.

Advantages of Larger VMs

Simpler Management: Fewer instances to monitor and maintain.

Higher Throughput per Instance: Larger VMs can handle bigger flows and more connections per instance.

Advantages of smaller VMs

Better Scaling: VMSS can scale out/in more granularly, matching traffic spikes.

Resilience: Failure of a single small VM has less impact than a large one.

Azure Load Balancing: More instances can improve distribution and redundancy