From: garciadeblas Date: Tue, 14 Jun 2016 07:26:09 +0000 (+0200) Subject: Multi-site NS X-Git-Url: https://osm.etsi.org/gitweb/?a=commitdiff_plain;h=617094b5dbab9c31be801a2f6ba6a513bd47f0f9;p=osm%2FFeatures.git Multi-site NS Signed-off-by: garciadeblas --- diff --git a/Release1/multi-site_network_services.md b/Release1/multi-site_network_services.md new file mode 100644 index 0000000..a5cf467 --- /dev/null +++ b/Release1/multi-site_network_services.md @@ -0,0 +1,90 @@ +# Multi-site NS - support of NS instances with VNFs running in different datacenters# + +## Proposer ## +**EUAG** + +## Type ## +**Feature** + +## Target MDG/TF ## +UI, RO, SO + +## Description ## +This feature request is about supporting NS instances with VNFs running in different datacenters. + +This feature assumes that the networks interconnecting several datacenters are pre-preprovisioned +(this pre-provision is out of the scope of this feature request). + +Current implementation in Release 0 allowed the deployment of a Network Service instance in the +datacenter chosen by the user. Specifically the MWC demo covered 4 deployments and the user decided +for every deployment the datacenter where to deploy. + +However, it is not possible to span an NS instance across different datacenters on a single +scenario (e.g. with some VNFs running in datacenter A and others in datacenter B). The workaround +of splitting the NS descriptor and deploy at different datacenters is not valid as it does not +allow to have common NS service primitives. + +There are several use cases where a multi-site NS applies, e.g.: + +- The deployment of a full core network with several PEs in different datacenters. +- The deployment of VoIP network where some elements are centralized in a datacenter while others +are distributed (e.g. SBC). + +In order to have this feature implemented, the following changes are devised: + +- In the UI: + - The end user should have the possibility to select for each VNF the specific datacenter where + it will be instantiated. + - If required, the end user should be able to select for each network in the NS the specific + datacenter where it will be instantiated. + - In case of networks spanning across several datacenters (e.g. an E-Line connecting a VNF in a + datacenter to another VNF in a different datacenter), the end user must select for each network + edge which one of the existing datacenter networks has to be used. +- In the RO: + - NS data models should differentiate between the data model for instances and the data model + for templates. The NS instance should allow to specify instance-related parameters for VNFs (in + this case, destination datacenter, but others can be added in the future, e.g. VNF instance + name). Additionally, the NS instance should allow to specify instance-related information for + networks: + - If the network is a single-DC network, i.e. it connects VNFs from the same datacenter, + there is no need to indicate the datacenter where the network will be deployed, since it + will be implicit from the VNFs. + - If the network is an inter-DC network, i.e. it connects VNFs from different datacenters, + the instance related information will include, for each network edge, the existing + datacenter network to be used. + - Database changes to annotate instance-related information, in this case the datacenter (or + datacenters) where it was deployed. + - Changes in northbound API to allow that information +- In the SO: + - NS data models should incorporate a data model for instances, different than the NS data + model for templates, where it should be possible to specify VNF instance-related parameters (in + this case, destination datacenter, but others can be added in the future, e.g. VNF instance + name). + - Connection to RO must be modified accordingly to the changes in RO northbound API. + - Database changes to annotate instance-related information, in this case the datacenter where + it was deployed. + +Specs for each module are to be written. For instance, in RO the spec is expected to include the +following devised changes: + +- openmano\_schemas.py: changes in schema "instance\_scenario\_create\_schema" to support the +specification of instance-related information per VNF and network. +- nfvo.py: changes in function "create\_instance" to support the new schema and take the actions +accordingly, including the appropriate calls to nfvo\_db functions. +- MySQL DB: changes in "instance\_scenarios", "instance\_vnfs", "instance\_vms" and +"instance\_nets" tables are required. Further investigation is required. +- nfvo\_db: changes in functions to update the previous tables accordingly. +- http\_server.py: changes in http\_post\_instances to take into account the changes in schemas, +nfvo functions and nfvo\_db functions. + +## Demo or definition of done ## +A test case consisting of the deployment of an NS instance with 2 interconnected VNFs (A and B) is +suggested. As a pre-requirement, there must be 2 datacenters (DC1, DC2) interconnected through a +pre-provisioned network visible at RO. As a convention, the network in DC1 will be named +DC1-interDC-net and the network in DC2 will be named DC2-interDC-net. + +At instantiation time, in a single transaction (e.g. launched from the GUI or the CLI), the NS +instance must be deployed specifying that VNF A is deployed in DC1 and VNF B in DC2, VNF A is +attached to DC1-interDC-net, and VNF B is attached to DC2-interDC-net. + +Information in both RO and SO should be read after the deployment to check validity of the test.