With a full mesh, the maximum number of switches is bound to the number of ports each switch can offer

With full mesh every switch have a direct connection to the other switches.If I have 3 switches, each switch will have at least 1 path to any other switch. This means you have 2 ports on every switch that you are using to form the mash. Obviously the maximum number of switches in a full mash of ds32b2 switches is .. 33 .. thus giving you a huge mesh of .. nothing since you are using all ports for ISLs and you have nothing left for hosts/storages. Now it's up to you to scale down and "steal" ISL ports to be used for hosts/storages .. Less ISLs means more hosts/storages.

You have also options .. if you need more ISLs between any two switches (in case you have a lot of traffic flowing from any two switches) maybe you have to use 2 (or more) ISLs between any two switches, thus cutting in half the maximum number of switches.

In a word.. full mesh is beautyfull but doesn't scale well.

With core/edge the math is simple and plain .... You have a core that will switch traffic between edges and storages. If your core have 64 ports and you need 20 ports for the storages, you have 44 free ports. With core/edge you need thick ISLs (2 or more between each edge and the core) thus giving you 22 edges for your beloved 64 ports core

Partial mesh is a subcase of the full mesh .. do your math