Communication infrastructure (CI) in microgrids (MGs) allows for the application of different control architectures for the secondary control (SC) layer. The use of new SC architectures involving CI is motivated by the need to increase MG resilience and handle the intermittent nature of distributed generation units (DGUs). The structure of secondary control is classified into three main categories including centralized SC (CSC) with a CI, distributed SC (DISC) generally with a low data rate CI, and decentralized SC (DESC) with communicationfree infrastructure. To meet the MGs operational constraints and optimize performance, control and communication must be utilized simultaneously in different control layers. In this survey, we review and classify all types of SC policies from CI based methods to communication-free policies, including: CSC, averaging based DISC, consensus-based DISC methods, containment pinning consensus, event-triggered DISC, washoutfilter based DESC, and state-estimation based DESC. Each structure is scrutinized from the view point of the relevant literature. Challenges such as clock drifts, cyber-security threats, and the advantage of event-triggered approaches are presented. Fully decentralized approaches based on state-estimation and observation methods are also addressed. Although these approaches eliminate the need of any CI for the voltage and frequency restoration, during black start process or other functionalities related to the tertiary layer a CI is required. Power hardwarein-the-loop (PHiL) experimental tests are carried out to compare the merits and applicability of the different SC structures.