[8023-POEP] March 06 802.3at Classification Ad Hoc: Final Summary

["Clay Stanford (LTC)" <claystanfordltc@xxxxxxxxx>]

FINAL SUMMARY OF THE IEEE 802.3at CLASSIFICATION AD
HOC MEETING

See formated version attached.


4 meetings via teleconference ending 3/2/06.

The following summary provided by Clay is based solely
on Clay’s perception of the feelings in the “room”. 
Classification requirements are divided into three
categories: Strong Agreement, Weak Agreement, and
Significant Controversy.  We will discuss these items
during the Denver meeting.

IEEE 802.3at CLASSIFICATION REQUIREMENTS

Strong Agreement
1.	Backward compatibility with 802.3af is required. 
2.	Mutual identification is required:
a. An AT-PD must be able to distinguish between AF-PSE
and AT-PSE 
b. An AT-PSE must be able to distinguish between AF-PD
and AT-PD. 
3.	Class policing will remain optional.
4.	25K signature resistance will not be changed.
5.	802.3af class resolution is too coarse and finer
resolution will be implemented.
6.	“Variable classification resolution” by either PSE
or PD will not be used.
7.	Adding more information into classification such as
vender ID will not be supported.
8.	One of the purposes of classification is to
implement power allocation (management) prior to
powering the PD.

Weak Agreement
1.	Classification power range is 2W to 100W. 
2.	Advanced power management, for example dynamic
power allocation will not be done in layer 1.  (It may
be performed in Layer 2, but that is outside the scope
of this ad hoc.)
3.	Layer 2 power management is optional.  
4.	Classification is mandatory in AT-PD and AT-PSE. 
5.	Classification method will support midspan and
endpoint PSEs, i.e. performed in layer 1.

Significant Controversy
1.	The classification scale should be roughly
Logarithmic, or log like. 
a.	Certainly logarithmic for high power classes 
b.	Probably linear for low power classes 
2.	The number of classes should be in the approximate
range of 30-40.
3.	One of the purposes of classification is to
implement class policing (i.e. current limit that
adjusts with class). 



ISSUES NEEDING FURTHER DISCUSSION
1.	What should the low end of the power range be; 1W,
2W, or other?
2.	Should we use worst-case or statistical analysis to
calculate utilization, and the number of classes
required? 
3.	What method should be used to implement
classification?
4.	Is it acceptable to power for example a 20 watt PD
using all 4-pair?  (In this example, it is assumed
that a 2-pair system can power above 20 watts.)
5.	Should a 4-pair PD that fails to get power provide
user with a two-level failure indication, one for an
AF-PSE and another for a 2-pair AT-PSE?
6.	How do the proposed requirements affect system test
time?
7.	How do the proposed requirements affect system test
complexity?
8.	Do the proposed requirements provide a good balance
between cost and benefit?
9.	Several questions interrelated to the architecture
used for 4-pair systems:
a. Should 4P verification be done during
classification or detection? 
a.	Should the PD have one signature (visible on all 4
pairs) or 2 separate signatures (one on Alt-A and the
other on Alt-B)?
b.	Should 4-pair systems be treated as two autonomous
2-pair systems?
c.	In a split cable installation using a 4P AT PSE, is
it expected that both PDs should receive power?



ITEMS TO CONSIDER WHEN EVALUATING VARIOUS
CLASSIFICATION METHODS
1.	Does the method meet all the requirements?
2.	How does the method affect system test time?
3.      How does the method affect system test
complexity?
4.	What is the PSE cost?
5.	What is the PD cost?

PRESENTATIONS
Yair Darshen provided a Comparison Table.
Classification_methods_review___Comparision_table.pdf

Yair referenced several previous works
http://grouper.ieee.org/groups/802/3/poep_study/public/jul05/
Classification Resolution Requirements Analysis
darshan_2_0705.pdf
How power Management Reduces System Costs
darshan_2_0305.pdf
Classification Resolution Analysis
darshan_3_0705.pdf

Yair presented a new version of the time based
classification scheme.
Backward_Compatible_Enhanced_Class_Rev_2.doc (34k

Yair presented an analysis of power supply
utilization.
Classification_Worst_case_Analysis.pdf

----------------------------
Steve Robbins presented “Extended Detection Protocol
for 4p PSE”
 4P_Detection_B_W.pdf (65k)

Steve presented a worst case power supply efficiency
analysis
WC_Analysis_2.pdf (300k)
----------------------------
Stanford
Presented issue with 4P PSE and AF PD.
4P_AT_PSE_with_AF_PD.pdf (43k)
Clay presented several possible 4P architectures.
4P_AT_PD.pdf (94k)
----------------------------
Christian Beia
Backwards compatible enhanced classification scheme
rev 1 and 2.



__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

313068510-Summary_2_23_06.doc