Distributed Controller Placement in Software-Defined Networks with Consistency and Interoperability Problems

Table 4

Comparison of related works on dmCP consistency.

Ref	Consistency aspect/properties			QoS metrics optimized	Method/consensus model	Weakness
	Connectivity	Capacity	Policy
[43]	X	√	√	Consensus/convergence time	Weak consistency with simple coordination layer	Higher conflicting tendency in the distribution of state updates to all nodes
[55]	X	√	X	Load balance	Weak consistency
[57]	X	X	√	Overhead	Load variance (LVS)
[58]	√	X	X	Forwarding loop	Greedy, combinatorial dependencies	Correctness issues
[59]	√	X	√	Memory, inconsistency	Incremental 2-phase commit	Increased packet header tagging overhead
[11, 60]	√		√	Violation, inconsistency	2-Phase commit
[63]	X	√	X	Availability, scalability delay	Adaptive weak consistency using cost-based	Critical traffic separation overhead
[65]	√	X	X	Forwarding loops, congestion, & blackhole	Direct message passing and partial knowledge	Marking update decisions with partial information
[66]	√	X	X	Blackhole freedom	Rule replacement with default matching rule, greedy	TCAM constraint, a forwarding loop
[68]	√	X	√	Conflicting policy updates	2-Phase commit, replicated state machine	TCAM constraints, tagging overhead
[70–72]	√	√	X	Loop freedom, loss, congestion	Clock synchronization	Message overhead
[73]	X	X	X	Availability	Adaptive weak consistency	Application requirement separation overhead
[74]	X	√	X	Load balancing	Customizable consistency generator, shim layer	Architecture modification, customization overhead
[75]	X	√	X	Congestions, deadlines, loss	Customizable update planner	Customization overhead
[76]	X	√		Synergy AP and CP	Adaptive consistency, K-means	Incur customization overhead
[77]	√	√	X	Overhead, availability	Adaptive consistency	Adaptability overhead
[78]	X	√	X	Congestion, throughput	LP, splittable flow, flow migration splittable flow	Flow reassembling overhead, flow migration cost
[79]	X	√	X	Lost
[80]	X	√	X	Congestion, convergence
[81]	X	√	X	Congestion	Timed update, dependency graph, greedy	Policy consistency might be hurt
[82, 83]	√	√	X	Congestion, latency, memory	Timed update, dependency graph
[84]	X	√	X	Congestion, complexity &overhead	Node-ordering	Prohibit e of 2-phase common thus, there is a high tendency of mismatch rule
[85, 86]	√	√	X	Latency	Distributed data store	High memory requirements
[87]	√	X	√	Security, inconsistency	2-PhaseCommit	TCAM constraint, tagging overhead
[88]	√	X	√
[89]	X	X	√	Inconsistent update	All-or-nothing principles	Transitional delay
[90]	X	X	√	Failure	Publish/Subscribe model. Atomic transactions	Not integrated with any controller
[92, 93]	X	X	√	Per packet consistency (PPC)	GA: 2-phase commit and rule replacement (RR)	TCAM constraint, tagging overhead
[94]	X	X	√	Policies preservation	LTLF, RR	Transitional delay