2016년 한국통신학회 하계종합학술발표회
OPNFV 환경에서의 자동화된 클러스터링 서비스 매니저 김민식, 두트렁쏜, 김영한* 숭실대학교
[email protected],
[email protected], *
[email protected]
An Automated Clustering Service Manager in Open Platform Network Function Virtualization Kim Min Sik, Truong-Xuan Do, Kim Young Han* Soongsil Univ
요 약 Network function virtualization (NFV) defined by European Telecommunication Standard Institute (ESTI) provides network operators with agile new service deployment, capital and operational expenditure reduction. Open platform network function virtualization (OPNFV) is an open source requirement and integration project for integrating upstream projects to realize the NFV specification. In this paper, we present an automated clustering manager for OPNFV orchestration. This manager is used to support high availability for stateless network function on the automated manner.
Ⅰ. Introduction The network configuration and setup for Telco application are quite complicated. It is not easy to introduce new services in a short time. To cope with that problem, Network function Virtualization (NFV) [1] is one of promising solution that enables network operators dynamically introduce new network services and create, destruct and change configuration as guaranteeing agile and elastic service. Currently, open platform network function virtualization (OPNFV) [2] is an open source project which provides Telcospecific requirements and integrates upstream projects for cloud virtualization (Openstack) [3] and softwaredefined network controller (Opendaylight) [4]. OPNFV also integrates other open platform for management and orchestration (OpenMano, Tacker). In this paper, we present an automated clustering manager for virtual network functions to support high availability in OPNFV platform. This manager is used to support HA for stateless network function on the automated manner. This manager automates the deployment of cluster of homogeneous network functions in OPNFV by using predefined template. First, we explain about OPNFV and basic clustering service Openstack Senlin[5]. Next, we present about our detailed design and architecture of our automated clustering manager. Last, we conclude with future work.
Figure 1 shows the architecture for OPNFV. Currently, the latest version of OPNFV is Brahmaputra. Figure 1 highlights upstream components in Brahmaputra along with the community lab infrastructure, where users can test the platform in different environments and on different hardware. These upstream components include cloud platform Openstack, SDN controllers (OpenDaylight, ONOS), and virtualization platform (KVM). Orchestration and management software also covered in OPNFV.
Figure 1. OPNFV architecture
Ⅱ. Related work B.
A. OPNFV
OpenStack clustering service
Senlin is one of OpenStack projects which manage groups of homogeneous cloud objects such as nova-
1069
2016년 한국통신학회 하계종합학술발표회 servers, heat-stacks and bare-metals. The goal of Senlin project is to make orchestration of collections of similar objects easier. User can create a cluster using the profile encoded the information needed for node creation. User is able to add / delete /update policies (auto-scaling, load-balancing) to a cluster. If there is some triggers from external monitoring tools, Senlin engine can perform some appropriate actions to cluster based on predefined policies attached to that cluster.
the cluster. Senlin driver is used to communicate with Senlin api and senlin engine to invoke APIs for cluter management. In our architecture, ceilometer is used to monitor virtual resources and trigger some actions when some ultilization metrics are over threshold. When alarm is triggered by other OpenStack service, receiver is used to react to that alarm and call senlinengine to excute process defined in policies.
Ⅲ. Clustering Service Manager for VNF A.
Overview of automated clustering manager
Figure 2 shows the relation of open source projects and ESTI reference architecture. Currently Openstack is selected as candidate for Network function virtualization infrastructure (NFVI) and virtual infrastructure manager (VIM). Openstack has some basic projects in charge of basic functions of VIM such as Ceilometer for monitoring virtual resources, heat for orchestration, nova for virtualized processor, neutron for network, and Senlin for clustering service. At the level of Virtual network function manager (VNF Manager), there is Tacker in charge of life cycle management, VNF configuration. At the orchestration
Figure 3. Automated Clustering Manager architecture Ⅳ. Conclusion and future work Our future work will extend the current automated clustering service to support virtual network function defined ESTI and more standard template file (using TOSCA template) ACKNOWLEDGMENT Figure 2. Design of Automated Clustering Manager level, OpenMaNO and TackerSFC in charge of deploying a network service, service function chaining. We introduce a automated clustering service manager which have following functions: provide automated deployment for a node cluster, support high availability of virtual network function, support auto-scaling, VNF placement etc. This manager is automated based on template file. This manager communicates with Senlin APIs for clustering management. The automated clustering manager takes cluster description included nodes, clusters, policies, trigger events in a template file; it will translate user defined description into parameters that can be consumed by the Senlin engine.
This work was supported by Institute for Information & communications Technology Promotion (IITP) grant funded by the Korea government (MSIP) (No. B190-152012, Global SDN/NFV Open-Source Software Core Module/Function Development) 참 고 문 헌 [1] “Network Functions Virtualisation (NFV); Architectual Franework”, ETSI GS NFV 002 v1.2.1, December 2014. [2] https://www.opnfv.org/ [3] https://www.openstack.org/ [4] https://www.opendaylight.org/
B.
Architecture of automated clustering manager
[5] Senlin: https://wiki.openstack.org/wiki/Senlin
Figure 3 show the architecture of automated clustering manager for orchestration. This manager consists of two main components: Parser and Senlin driver. Parser is used to analyze the template files for input parameters such as profile definition, cluster size, policies, and receivers for making and managing
1070