Centralized versus distributed network control. (a) A multicell system managed by centralized resource controller. This controller processes all network information jointly. (b) A distributed multicell system requires no centralized control. Each cell performs resource allocation based on local channel knowledge (and possibly limited intercell information). 

Contexts in source publication

Context 1

... challenges. This includes among others the need for slot level synchroni- zation for large network areas. However, this problem may be partly alleviated by clustering the optimization. Another severe problem is the need for the joint processing of traffic and channel quality parameters fed back by all network nodes to a central control unit [see Fig. 1(a)], which necessitates significant computational power and huge signaling overhead. This makes global network coordination hard to realize in practice, especially in mobile settings where the control unit ought to cope with 1 The operating point is the level of SINR needed to operate on the link, below which the call may be dropped. fast ...

Context 2

... the approaches above, including greedy optimization, the need may exist for centralized knowl- edge of all channel and interference state conditions for all nodes in the network. In the case of the greedy approaches, the algorithm then only visits the cell virtually, and the actual computation takes place within the central control unit shown in Fig. 1(a). Centralized channel state information for a many-cell many-user network involves immense signaling overhead and will not allow the extraction of diversity gains in fast-fading channel components. To circumvent this problem, the design of so-called distributed resource allocation techniques is crucial. Distributed optimization refers ...

Context 3

... in the asymptotic number of users regime, and to some rare exceptions (such as [32]), the methods presented so far require CSI of all cells to be centrally collected and processed [ Fig. 1(a)]. As the network size grows, this turns into a serious impairment, especially in high-mobility scenarios where the coherence time of the channel gain may not exceed a few tens of milliseconds. In a distributed scenario [ Fig. 1(b)], a resource allocation unit is instead located in each cell, handling scheduling and resource management ...

Context 4

... (such as [32]), the methods presented so far require CSI of all cells to be centrally collected and processed [ Fig. 1(a)]. As the network size grows, this turns into a serious impairment, especially in high-mobility scenarios where the coherence time of the channel gain may not exceed a few tens of milliseconds. In a distributed scenario [ Fig. 1(b)], a resource allocation unit is instead located in each cell, handling scheduling and resource management for the cell users based on local CSI and possibly limited remote CSI. Note that local CSI includes any information that is observable by a local user, such as the gain from its preferred AP as well as the power from the various ...