Dynamic memory for Exchange 2013 – why not supported?

Recommendation why to not to use dynamic disks on crucial applications might be explained quite easily – since system is saving new stuff on the disk, each save operation equals disk expansion – which takes valuable time. But what is so special with Exchange and dynamic memory which usage is also not recommended by Microsoft?

Bhargav Shukla and Paul Robichaux in their book “Core Solutions of Microsoft Exchange Server 2013 (MCSE)” explain why dynamic memory might be a problem for Exchange 2013, here is the piece that is very interesting:

“Exchange 2013 code is optimized to strike a balance between the efficient use of memory and reducing the I/O footprint. To achieve these efficiencies, Exchange relies on a calculated cache for each database being hosted on the server, as well as the memory reserved for Exchange subsystems. When dynamic memory is in use, this can result in incorrect memory calculations and it can cause Exchange to start with less memory than is available.”

So that makes sense, and here is the place where RAM allocation is shown in a really nice way:


Also interesting thing is that oversubscription of processor is supported, the recommended ratio is 1:1 (of course ;]), but supported is 2:1, so that would mean that for ex. 1 physical CPU of an ESX host, can be shared only between two machines (Exchange or any other).