Gatekeeper Redundancy

There seems to be two different options for gatekeeper design, clustering and directory gatekeeper (DGK). It looks as if the preferred method is clustering gatekeepers together rather than creating a DGK. Using gatekeeper clustering, you configure a local gatekeeper at each site but then you make redundant alternate connection other sites. This allows for the gatekeeper to provide primary call routing for the main site along with providing alternate call routing for other locations connected to it and vice versa.

DGK depends on the Hot Standby Router Protocol (HSRP) or by configuring multiple DGK's in your environment. I'm still trying to wrap my head around how this particular way works a little more.