IEC Type 4: AR/VR oriented Edge Stack for Integrated Edge Cloud (IEC) Blueprint Family

Use Case Attributes

Description

Informational

Type

New Blueprint for VR/AR on the Network Edge

Blueprint Family

Integrated Edge Cloud (IEC)

Use Case

Deployment of generic edge end and cloud environment for VR/AR cloud streaming

Blueprint proposed Name

IEC Type 4: AR/VR oriented Internet Edge Stack for Integrated Edge Cloud (IEC) Blueprint Family

Initial POD Cost (capex)

NVIDIA RTX GPUs

Chelsio T580-CR NICs

Scale & Type

Generic blueprint PoC:

•   One master node and up to 5 worker nodes with mixed Linux and Windows OS
•   Each server, x86/ARM server with nVidia RTX GPUs (Titan or GeForce TBD)

Large scale deployment:

•  Number of servers, x86/ARM server or deep edge class, is site dependent (footprint)
•  vGPU and federation supported class, e.g. NVIDIA Tesla K80 GPUs;
•  Chelsio T580-CR NIC

Applications

Generic blueprint POD: Small scale cloud AR/VR rendering farm with generic SO.

Production/commercial service:

1. Consumer applications: High performance premium gaming, 3D video for movies, live concerts, events, LBE, etc.
2. Enterprise applications: training/education, product design collaboration, manufacturing, maintenance, data analytical etc,

Power and memory restrictions

N/A

Infrastructure orchestration

Docker 18.09.4 or above (19.03 may be needed to run windows container with nVidia GUP support) and K8s 1.14.1 or above- Container Orchestration, VMWare VM

OS - Ubuntu 18.04.2, windows server 2019

Under Cloud Orchestration - Airship v1.0 (TBD)

SDN

Calico and K8s, Memcached load balancer (optional)

Workload Type

VR and AR applications with split rendering runtime running inside Containers or VM

Additional Details

The test configuration consists of 3 machines connected using ethernet switch: a master and 2 worker nodes, each with TBD processor clocked at TBD GHz, with TBD GB of RAM and Ubuntu operating system for master, windows server 2019 or later for worker. MTU of 1450B is configured (to compensate for GTP tunnel header). Each windows server preconfigures with 2-3 VMs with fixed GPU allocation per VM.

Committer

Committer

Company

Committer Bio

Committer Picture

Self Nominate for PTL (Y/N)

Contributor

Contributor

Company

Contributor Bio

Contributor Picture

Self Nominate for PTL (Y/N)