Introduction:
The INF(infrastructure) project provides open source reference implementation of Edge Cloud infrastructure according to the O-RAN WG6 specification to be used with the other O-RAN OSC projects such as O-CU, O-DU and in the future potential O-RU to create a complete reference implementation of the different O-RAN use case scenarios as defined by O-RAN Alliance work groups. The work in the INF project will following “Open Collaboration”, “Open Design”, “Open Development” and “Open Source”.
Goals:
- When make sense use components / parts from existing open source project that have wide and diverse community support and adoption
- Provide Linux distribution that will be used as the foundation for the Edge Cloud which meet the small footprint and real-time performance requirements required by the O-CU and especially O-DU use cases, and in the future O-RU.
- Provide Edge Cloud implementation which support both Kubernetes/Container and OpenStack/VM which able to support deployment of small scale through large scale Edge Clouds, required to support the different use case scenarios, and all while providing cost efficient solution.
- Provide Edge Cloud which support accelerator abstraction layer allowing decoupling between the application that runs on the Edge Cloud that requires the accelerator and the actual physical accelerator.
JIRA: https://jira.o-ran-sc.org/projects/INF
PTL:
- Xiaohua Zhang (xiaohua.zhang@windriver.com)
Committers:
- Jackie Huang (Jackie.huang@windriver.com)
- Jianqiang Lu (lujq1@lenovo.com)
- Xiaohua Zhang (xiaohua.zhang@windriver.com)
Contributor:
- Bin Yang (bin.yang@windriver.com)
Weichen Ni (niweichen@chinamobile.com)
- Xiaogang Yan (yanxiaogang@chinamobile.com)
- Kevin Zhang (zhangnianhua@inspur.com)
EPICs in Release A
O-RAN SC Ver A SW Requirements 20190731a.docx
Title | [INF-A-F01] Akraino blueprint for RIC |
Description | As a Service Operator I need a REC blueprint in order to deploy RIC platforms |
Acceptance Criteria | Blueprint executes and platform is available for RIC deployment |
Source | OSC “A” (AT&T) |
Title | [INF-A-F02] VNF/CNF Orchestration |
Description | As a Service Operator I need to deploy virtual or container based applications utilizing pre-O1* interfaces based on ETSI MANO Or-Vi and Vnfm-Vi interfaces to a cloud platform. This implementation will demonstrate how to use the exposed O1* Orchestration services for orchestration between Service/VNF and Cloud Orchestration. |
Acceptance Criteria | Using StarlingX Cloudify deploy a VNF or CNF and demonstrate Start/Stop/Reboot lifecycle commands based on FM collection using StarlingX FM from the deployed application as a pre-specification O1 interface(O1). |
Source | OSC “A” (Inspur) |
Title | [INF-A-F03] Acceleration Abstraction Layer (AAL) |
Description | O-CU and O-DU are typically implemented as software running in VMs or containers. Many functions in the software stack are compute-intensive, latency-sensitive, and also power-consuming when running on general-purpose CPUs. Examples include physical layer functions and crypto algorithms in the PDCP. A general trend is to offload these functions to hardware accelerators (e.g., in the form of PCI-E devices). There are many types of hardware accelerator, e.g., using ASIC, NPU, FPGA, and so on. To allow full decoupling of the O-CU and O-DU software from the acceleration hardware, this project defines and implements an Acceleration Abstraction Layer (AAL), which hides the hardware differences and presents a unified API to O-CU and O-DU software. Details at https://wiki.o-ran-sc.org/download/attachments/3604609/LNV-2019.06.19-OSC%20AAL.pptx |
Acceptance Criteria | Platform is available for O-CU and O-DU deployment. |
Source | OSC “A” (Lenovo) |
Title | [INF-A-F04] Real time platform to deploy the O-CU and O-DU |
Description | In O-RAN architecture, the O-DU and O-CU could have different deployed scenarios. In general the performance sensitive parts of the 5G stack require real time platform, the platform should support the Preemptive Scheduling feature. For example, from implementation perspective, the non-virtualized DU, VM based DU and container based DU are requiring a real time host system. Following requirements are going to address the container based solution: 1. Support Node Feature Discovery 2. Support CPU Affinity and Isolation 3. Support Dynamic HugePages Allocation 4. Support Topology Manager 5. Support Device Plugin And for the network requirements, the following should be supported: 1. Multiple Networking Interface 2. High performance data plane including the DPDK based vswitch and PCI pass-through/SR-IOV. (The details of requirement could be found in https://oranalliance.atlassian.net/wiki/download/attachments/106233916/ATT-2019.%20O-RAN%20WG6%20CAD.docx?api=v2 ) |
Acceptance Criteria | Platform is available for O-CU and O-DU deployment |
Source | OSC “A” (Wind River) |
Recent space activity
