Cologne: A Declarative Distributed Constraint Optimization Platform

• constraint optimization 2

http://repository.upenn.edu/cgi/viewcontent.cgi?article=1720&context=cis_papers

Cologne integrates Gecode, an off-the-shelf constraint solver, and RapidNet declarative networking engine for communicating policy decisions among different solver nodes.

• Cologne allows cloud providers to formally model cloud resources and formulate orchestration decisions as a COP given goals and constraints.

  RapidNet: an incremental bottom-up distributed Datalog evaluation engine

  Gecode: a top-down goal-oriented constraint solver

• centralized deployment mode: takes as input system states gathered from all nodes in the network, and a set of policy constraints and goals specified using Colog declarative language. These specifications are then used by a constraint solver to automatically generate optimization command. These commands are then input into each node’s configuration layer, to generate physical operations to directly manipulate resources at each node.

• decentralized mode: multiple Cologne instances, typically one for each node. … each node has a set of neighbor nodes that it can directly communicate with. A distributed query engine coordinates the exchange of system states and optimization output amongst Cologne instances, in order to achieve a global objective.

cloud orchestration involves the creation, management, manipulation and decommissioning of cloud resources, including compute, storage and network devices, in order to realize customer SLAs, while conforming to operational objectives of the cloud service providers.