Skip to main content
RISE logo
OCP edge node half-rack selfcontained

An OCP based refurbished half-rack edge node

26 April 2022, 09:23

This OCP edge node project set out to maximize the environmental responsibility and at the same time serve the needs for an enterprise 5G SA edge node deployment. The project should enable a maximum number of edge application users with a small footprint for the edge node with an easy to use requirement from the enterprise. The price point should also meet a low target cost for the complete node.

The basic idea was to design a recycling product out of a regular OCP rack and OCP gear into a low-cost high-capacity edge node for industry and smart city use cases. The design should also be self-contained and with plug-and-play characteristics.

The sustainability was addressed by maximizing the server infrastructure re-use by cutting an OCP rack in half, contain it and re-use OCP servers, switching and UPS in a second-life application. Also a holistic cooling concept was used taking control of server fans and cooler fans to minimize the energy use. The edge node was made self-sustainable requiring only an Ethernet connection and a 16A plug to meet the plug-and-play requirement. 15 inexpensive re-used servers with added GPU cards make the edge node to a compact powerful compute device with a competitive price to be directly connected to the 5G radio base station.

Edge is a wide definition including device edge on smart devices such as smartphones, gateways, drones etc. all the way to the Internet edge datacenters of fairly large sizes, in the 1 MW range, at Internet connect points. On the devices the edge is mostly battery and compute resource limited. This can be improved with capacity at the edge of the last mile network. Lately with the low latency improvements in 5G networks it makes sense to put compute capacity in the network close to the base station to off-load battery and compute limited devices and move content or compute even closer to the end-users.

This project was driven by the needs from 5G network edge use cases. More specifically a project for implementing a network of edge nodes connected to 5G base stations in a testbed on campus at Luleå university of technology and for testbeds planned in a mine, in a smart city environment and on an airport. Edge here is the edge of a 5G network sometimes defined as deep network edge. The design must fit a traffic case for a macro 5G base stations serving a fairly large area.

The objectives of the design of the edge node was to be in the size of 10-15 servers with GPU capacity, possible to connect with one 16 A connector and an Ethernet connection-only with plug and play characteristics. The objective was also to show how to recycle regular server infrastructure in an edge node. Also if needed the cooling system should be possible to connect to a building heating network. Back-up power and UPS was to be provided within the cabinet itself. The energy efficiency of the node is also important by co-operate the server fans with the “facility” fans. The holistic view on controlling the performance and energy efficiency is a key component.

The design with a half recycled OCP rack with 15 Leopard OCP servers met the improved IT infrastructure life-time requirements. The capacity of 15 CPU servers with added GPU capacity would enable service for hundred of users in a high-density macro 5G site. The integration of UPS and cooling enabled a plug-and-play scenario. The cost of re-used OCP servers enable a low price point.

 

Tor Björn Minde
CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

* Mandatory By submitting the form, RISE will process your personal data.

Jonas Gustafsson

Jonas Gustafsson

Research program manager

+46 70 333 00 35

Read more about Jonas

Contact Jonas
CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

* Mandatory By submitting the form, RISE will process your personal data.

2024-04-11

2024-04-05

2024-04-02

2024-04-01

2024-03-10

2024-02-21

2024-01-30

2024-01-17

2024-01-14

2024-01-03

2023-12-22

2023-12-20

2023-12-18

2023-12-12

2023-12-09

2023-11-30

2023-11-28

2023-11-20

2023-11-17

2023-11-12

2023-11-05

2023-10-29

2023-10-24

2023-10-22

2023-10-10

2023-10-08

2023-10-06

2023-09-22

2023-09-20

2023-09-13

2023-08-29

2023-08-28

2023-08-16

2023-08-07

2023-07-19

2023-06-30

2023-06-26

2023-06-23

2023-06-05

2023-06-02

2023-05-16

2023-05-05

2023-04-26

2023-04-26

2023-04-24

2023-04-22

2023-04-22

2023-04-12

2023-04-03

2023-03-31

2023-03-24

2023-03-08

2023-02-28

2023-02-27

2023-01-29

2023-01-25

2023-01-18

2023-01-16

2022-12-22

2022-12-20

2022-12-14

2022-11-18

2022-10-21

2022-10-12

2022-10-10

2022-10-07

2022-10-05

2022-10-01

2022-09-26

2022-09-20

2022-09-15

2022-09-15

2022-09-15

2022-09-14

2022-09-12

2022-09-09

2022-09-06

2022-08-31

2022-06-30

2022-06-05

2022-06-01

2022-05-31

2022-05-29

2022-05-22

2022-05-17

2022-05-13

2022-05-04

2022-04-26

2022-04-13

2022-04-08

2022-04-07

2022-04-06

2022-04-05

2022-03-28

2022-03-25

2022-03-16

2022-02-25

2022-02-23

2022-02-22

2022-02-18

2022-02-16

2022-02-15

2022-02-10

2022-01-28

2022-01-18

2022-01-13

2022-01-05

2022-01-03

2021-12-30

2021-12-29

2021-12-23

2021-12-21

2021-12-18

2021-12-15

2021-12-14

2021-12-09

2021-12-02

2021-11-24

2021-11-19

2021-11-18

2021-11-16

2021-11-10

2021-11-08

2021-11-04

2021-11-03

2021-10-19

2021-10-16

2021-10-15

2021-10-10

2021-10-10

2021-10-08

2021-10-07

2021-10-06

2021-09-14

2021-08-23

2021-08-19

2021-07-10

2021-07-08

2021-07-07

2021-07-06

2021-07-05

2021-06-30

2021-06-28

2021-06-23

2021-06-13

2021-06-07

2021-06-07

2021-06-05

2021-06-02

2021-05-30

2021-05-06

2021-04-20

2021-03-27

2021-03-21

2021-03-10

2021-03-10

2021-02-11

2021-01-15

2021-01-14

2021-01-08

2021-01-07

2021-01-04

2020-12-30

2020-12-30

2020-12-28

2020-12-18

2020-12-11

2020-12-11

2020-11-28

2020-11-26

2020-11-25

2020-11-20

2020-11-20

2020-11-16

2020-11-15

2020-11-10

2020-11-05

2020-11-04

2020-10-22

2020-10-21

2020-10-08

2020-10-05

2020-10-02

2020-09-30

2020-09-24

2020-09-17

2020-09-11

2020-08-31

2020-08-10

2020-07-07

2020-07-06

2020-07-05

2020-07-05

2020-07-03

2020-07-01

2020-06-30

2020-06-29

2020-05-29

2020-05-11

2020-04-20

2020-04-13

2020-03-28

2020-02-10

2020-01-29

2020-01-17

2019-12-20

2019-12-20

2019-12-17

2019-12-06

2019-11-26

2019-11-18

2019-10-25

2019-10-11

2019-09-11

2019-09-04

2019-08-27

2019-08-22

2019-08-13

2019-08-01

2019-07-29

2019-07-26

2019-07-26

2019-06-27

2019-06-26

2019-06-05

2019-06-05

2019-05-24

2019-05-09

2019-05-08

2019-05-08

2019-04-10