A scale up production deployment is where all of the Enterprise Server for .NET roles ( Listener, Dispatcher , Monitor and SEP) are hosted on a single machine, typically an enterprise class blade. As more resource is required (disk, CPU, memory), it is added to the machine. The problem with scale up architecture is that once the upgrade capacity of the single machine is reached, a newer machine with more capacity needs commissioning, with the associated downtime and costs.
A scale out deployment is where each role can be hosted on multiple, cheaper machines, or even in a cloud deployment. As more capacity is required, e.g. SEP roles, then more machines can be added to the deployment to host more SEPs. When the extra resource is no longer required, they can be removed.
If network load balancers (NLBs) are used between the roles, introducing and removing these machines need not prevent work continuing on the remaining infrastructure. The use of an NLB will also ensure that if a machine or role is removed from the deployment, the remaining machines will be given more share of the load.
Enterprise Server for .NET architecture was designed with scale out in mind and should be thought of as a scale out solution . It will work in a scale up scenario, but the advantages of scale out would be lost.
Please refer to the Scaling out topic for further information.