1.
energy aware equipments measurement method Ericsson’s contribution to ECONET project
march 2012

2. agenda Actual EE Measurement Methods The Traffic Load Gradient
Traffic Load Multi-Levels
Energy Aware Measurement Method
Conclusions

3. Actual EE Measurement Methods
According to actual Standardization bodies, Energy Efficiency shall be determined by measuring the EUT at few fixed traffic load states, such as for instance per ATIS “Measurements methodologies”:
TEER = Max Demonstrated Throughput / Weighted Power
Max Demonstrated Throughput is the highest achievable system throughput at Non-Drop Rate (bps).
Weighted power is Pw = a*Pu1(%) + b*Pu2(%) + c*Pu3(%) where:
(a, b, c)  = power weighting at each utilization level (a+b+c=1)
(Pu1, Pu2, Pu3) = power consumption at the specified system % of Traffic Load level, 100%, 50%, 0%.

4. Actual ee measurement methods
Actual EE measurement methods lack of:
clear indication of the trade-off between Power States Transition vs. QoS
slavishly approximate real traffic shape
not allowing to highlight and exploit all the Power Management features of the EUT (Equipment Under Test)

5. Improved ee measurement method
Improvement to EE Measurement Method is proposed, based on
Traffic Load Gradient (TLG)
Traffic Load multi-Levels (TLL)

6. Traffic load gradient
When the Power States dynamically change, if the EUT does any dynamic Power Management to automatically adapt power consumption to the actual rate, the transition between the leaving and the entering Power States should be guaranteed at zero packet loss
A new test function is herein introduced to characterize the EUT under this perspective, namely the Traffic Load Gradient (TLG) function of the Traffic Load variation entity and rapidity

7. Traffic load gradient
Traffic loss during Power States transition shall be verified by applying linear gradients of traffic from the estimated maximum traffic load of the given lower Power State to the minimum to enter the next given higher Power State (Lmin  Lmax; d1) and from the minimum traffic load of the given higher Power State to the estimated traffic load to enter the next given lower Power State (Lmax  Lmin; d2) in the given time.
If during transitions traffic loss is verified, the design shall be arranged so to avoid such event, or gradient magnitude and/or time duration should be relaxed e.g.: from fast gradient to slow gradient.
If it is not possible getting zero packet loss transition, the “no-error-free condition” during transitions between the given Power States should be declared. As such spontaneous transition between the given Power States may be inhibited by the Operator, according to the Service Level Agreement.

8. Traffic load gradient
Actually there is not any Traffic Load Gradient requirement or specification available, nor at Standardization level neither in literature
Physical measurement on live Network may be required to derive realistic Traffic Load Gradients set (fast, slow)
weekends
daytime
night-time
sudden peak

9. Traffic load Multi-levels
According to most demanding actual Standards, Traffic Load Levels (percentage of traffic load) to be considered in the measurement are typically are just: 100%, i.e. max traffic load; 50%, i.e. half traffic load; 0%, i.e. no load
When it comes to the new paradigm of Energy Aware devices specifically designed to dynamically adapt power consumption to actual traffic, few samples of Traffic Load Levels looks to be inadequate
An effective Measurement Method should as much as possible slavishly follow the real traffic shape so to fully characterize Energy Aware EUT power management capabilities, by means of Traffic Load multiple Levels (TLL)

10. Traffic load Multi-levels
Much finer discrimination is required, approximating as much as possible the real traffic shape to fully exploit the EUT Power Management features
Instead of a few subset of Traffic Load levels, the Measurement Method shall comprise at least steps of 10% traffic load levels from 100% to 0%, going to explore also the various no-traffic conditions such as Fast-Sleeping, Deep-Sleeping and Stand-by.
As such TLG level can be defined as 10% of max traffic capacity, while TLG timing is tbd
Actual
Suggested

11. Energy aware Measurement Method
Effective Energy Aware equipments Measurement Method shall be based on slavish traffic load shaping
Traffic Load Gradient and Traffic Load multi-Levels shall be defined
The power consumption at any of the Traffic Load multi-Levels shall be measured, by verifying the free errors transition between each level when the given Traffic Load Gradient (rising and falling) is applied
The Equipment’s Energy Efficiency Ratio shall be defined as (weights are tbd):
∑ (weighted max Power at each TLL) αP100%+βP90%+ … +σP0%+ρPFS+ φ PDS+ ωPSTBY
EEER = =
∑ (weighted max Traffic Load at each TLG) αT100% +βT90% + … +σT0%
EEERFAST will apply for fast TLG; both rising and fall traffic load transitions shall be tested; if errors are observed during the test, EUT will be declassed to
EEERSLOW applying for slow TLG, if errors persist EUT will be declassed to
No spontaneous PMM

12. conclusions
Several Power Management Modes (power-states) may apply at an Energy Aware device as function of the traffic load, plus additional status such as Fast Sleeping, Deep Sleeping, Standby
To optimize the effectiveness of the measurement method, suitable Traffic Load Shaping (namely TLG and TLL) shall apply so to
fully exploit the EUT power saving (Power Management Modes) capabilities
guaranty QoS by verifying zero traffic loss towards the gradient’s rising and falling directions
push for the development and integration of Power Management features approximating as closest as possible the real traffic shape

13. conclusions
Ericsson requests that the development of a novel and effective Energy Aware Equipments Measurement Method based on TLG and TLL as herein briefly described is included in the scope of the ECONET project and proposed to relevant Standardization bodies
Part of the effort will be the TLG timing and EEER weightings definition