Presentation is loading. Please wait.

Presentation is loading. Please wait.

Energy Audit aaS with OPNFV

Similar presentations


Presentation on theme: "Energy Audit aaS with OPNFV"— Presentation transcript:

1

2 Energy Audit aaS with OPNFV
B.Hérard, Orange M.Richomme, Orange

3 Power efficency is a key driver for Telco industry

4 Note that studies showed that virtualization could lead to an increase of power consumption and is still a matter of debate

5 Today, the Orange ITN (Information Technologies & Networks) factory is made up of around 100,000 mobile access points, dozens of data centres, and service platforms. Together, they generate 75% of the Group’s total CO2 emissions and account for 80% of its energy consumption. The potential for improvements is huge. 500 GWh. That’s how much energy Orange’s data centres consume in a year. The average annual consumption of 75,000 French households. That figure is not alarming in itself. Yet. But it will increase in years to come as pressure from data usage grows. More than ever, the “greening” of data centres is a strategic issue. Orange has committed to reduce its greenhouse gas emissions by 20% between 2006 and This is a real challenge within the context of rapidly growing usage and the relentless increase in data traffic. Src:

6

7 Energy Efficiency is a key driver….
However it is poorly covered by Open Source communities…

8 Energy Audit aaS in OPNFV
Collect: get real time Power consumption of Pharos PODs (all the nodes) Corelate: leverage determinist CI runs to establish Energy profiles Optimize: implement policy or placement rules to optimize power consumption

9

10

11 Power collection IPMI Redfish
Src: tests done in Orange Labs on DELL server, to be confirmed on other servers

12 Power collection Build the collector Install it on your jumphost
Need access to IPMI/Redfish Need Internet access to push measures to InfluxDB

13 Pull Down Functest Image Pull Down Yardstick Image
Docker Hub Docker Hub During CI Run Instructions trigger Energy measurement Push result to Influx DB Pull Down Functest Image Pull Down Yardstick Image Energy API Functest Yardstick Pharos POD (SUT) Collector

14 Python Power consumption SDK
Indicate (using decorators) the cases you want to monitor You can precise start/stop or intermediate events (e.g. onap deployed) @energy.enable_recording def run(self): Energy.energy.step("step1") self.do_some_stuff1() Energy.energy.step("step2") self.do_some_stuff2() @energy.enable_recording def run(self): self.do_some_stuff1() self.do_some_stuff2()

15 Vizualization

16 Vizualization

17 Next steps Opportunity for a dedicated « Energy » project? (initial work triggered during the Plugfest in Paris + bootstrap with Functest) Understand the different phases, and use the event API to visualize steps Get Data, get data, get data ….need data to validate test/VNF profiling => sync with Bamboo

18 Next steps Statistics & big Data….
Link with Bamboo (OPNFV implementation of PNDA) Src: post processing on yardstick results done by Alassane Samba (PHD student in Orange)

19 Conclusions Power consumption monitoring initiated with an end to end prototype up&running Exciting but still challenging (not sure the profiles we are collecting will be consistent/acurate)…empiric way… New area for POD optimization study (inputs for watcher?) contact benoit.herard AT orange.com

20 谢谢 - Thank you - Merci


Download ppt "Energy Audit aaS with OPNFV"

Similar presentations


Ads by Google